Free
Perioperative Medicine  |   June 2018
Association of Polypharmacy with Survival, Complications, and Healthcare Resource Use after Elective Noncardiac Surgery: A Population-based Cohort Study
Author Notes
  • From the Departments of Anesthesiology and Pain Medicine (D.I.M., G.L.B.) and Internal Medicine (C.v.W.) and the School of Epidemiology and Public Health (D.I.M., C.v.W.), University of Ottawa, Ottawa, Ontario, Canada; the Departments of Anesthesiology and Pain Medicine (D.I.M., G.L.B.) and Internal Medicine (C.v.W.) and the Research Institute (D.I.M., G.L.B., C.v.W.), The Ottawa Hospital, Ottawa, Ontario, Canada; and the Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada (D.I.M., C.A.W., C.v.W.).
  • This article is featured in “This Month in Anesthesiology,” page 1A.
    This article is featured in “This Month in Anesthesiology,” page 1A.×
  • Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are available in both the HTML and PDF versions of this article. Links to the digital files are provided in the HTML text of this article on the Journal’s Web site (www.anesthesiology.org).
    Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are available in both the HTML and PDF versions of this article. Links to the digital files are provided in the HTML text of this article on the Journal’s Web site (www.anesthesiology.org).×
  • Submitted for publication June 14, 2017. Accepted for publication January 9, 2018.
    Submitted for publication June 14, 2017. Accepted for publication January 9, 2018.×
  • Address correspondence to Dr. McIsaac: Ottawa Hospital, 1053 Carling Avenue, Room B311, Ottawa, Ontario K1Y 4E9, Canada. dmcisaac@toh.ca. Information on purchasing reprints may be found at www.anesthesiology.org or on the masthead page at the beginning of this issue. Anesthesiology’s articles are made freely accessible to all readers, for personal use only, 6 months from the cover date of the issue.
Article Information
Perioperative Medicine / Clinical Science / Pharmacology
Perioperative Medicine   |   June 2018
Association of Polypharmacy with Survival, Complications, and Healthcare Resource Use after Elective Noncardiac Surgery: A Population-based Cohort Study
Anesthesiology 6 2018, Vol.128, 1140-1150. doi:10.1097/ALN.0000000000002124
Anesthesiology 6 2018, Vol.128, 1140-1150. doi:10.1097/ALN.0000000000002124
Abstract

Background: Polypharmacy is increasingly prevalent in older patients and is associated with adverse events among medical patients. The impact of polypharmacy on outcomes after elective surgery is poorly described. The authors’ objective was to measure the association of polypharmacy with survival, complications, and resource use among older patients undergoing elective surgery.

Methods: After registration (NCT03133182), the authors identified all individuals older than 65 yr old having their first elective noncardiac surgery in Ontario, Canada, between 2002 and 2014. Using linked administrative data, the authors identified all prescriptions dispensed in the 90 days before surgery and classified people receiving five or more unique medications with polypharmacy. The associations of polypharmacy with 90-day survival (primary outcome), complications, length of stay, costs, discharge location, and readmissions were estimated after multilevel, multivariable adjustment for demographics, comorbidities, previous healthcare use, and surgical factors. Prespecified and post hoc sensitivity analyses were also performed.

Results: Of 266,499 patients identified, 146,026 (54.8%) had polypharmacy. Death within 90 days occurred in 4,356 (3.0%) patients with polypharmacy and 1,919 (1.6%) without (adjusted hazard ratio = 1.21; 95% CI, 1.14 to 1.27). Sensitivity analyses demonstrated no increase in effect when only high-risk medications were considered and attenuation of the effect when only prescriptions filled in the 30 preoperative days were considered (hazard ratio = 1.07). Associations were attenuated or not significant in patients with frailty and higher comorbidity scores.

Conclusions: Older patients with polypharmacy represent a high-risk stratum of the perioperative population. However, the authors’ findings call into question the causality and generalizability of the polypharmacy-adverse outcome association that is well documented in nonsurgical patients.

What We Already Know about This Topic
  • General medical literature has established an association between polypharmacy and adverse medical outcomes

  • It is unclear whether surgical patients with polypharmacy are also at increased risk of adverse outcomes

What This Article Tells Us That Is New
  • Polypharmacy is associated with increased postoperative adverse events

  • The association is tenuous, may be limited to specific patient groups or medication types, and may be a marker for disease burden

  • Further study is necessary before any clinical practice changes can be considered

THE aging of our population has a substantial impact on surgical utilization and outcomes. Individuals older than 65 yr old have surgery at a significantly higher rate than younger age groups,1  and advanced age is associated with a two- to fourfold increase in adverse postoperative events and increased healthcare resource use.2,3  Accordingly, improving the outcomes of older surgical patients has been identified as 1 of the 10 most important priorities in perioperative research by patients and clinicians.4 
Although age is an independent predictor of adverse postoperative outcomes, it is an unmodifiable risk factor that likely acts as a proxy measure for other conditions, such as multimorbidity and frailty, that accumulate across the lifespan. In conjunction with the development of chronic illnesses and decreased function associated with aging, polypharmacy is common in older individuals. Polypharmacy, typically defined as taking more than four to six medications concurrently,5,6  has doubled in prevalence in the past 20 yr among older community-dwelling individuals.7  In nonsurgical populations, a recent systematic review found that polypharmacy was associated with increased risk of falls, poor functional status, increased risk of hip fracture, adverse drug events, increased rates of unplanned hospital admissions, poor self-perceived health status, and increased mortality.8  However, patients who take multiple medications may also be sicker at baseline, which puts the polypharmacy–outcome association at risk of confounding bias.
The epidemiology of polypharmacy in the perioperative setting is poorly described. Single-center studies of general surgery patients suggest that polypharmacy is associated with a three- to fourfold increase in the odds of serious complications.9,10  Hip fracture patients with polypharmacy have an increased risk of readmission11  and often are discharged on the same medications that may have contributed to the fracture.12  The prevalence of polypharmacy in elective surgery patients and the impact that polypharmacy may have on population-level postoperative outcomes have not been described. Strong evidence of an unbiased association between polypharmacy and adverse outcomes could inform strategies to address polypharmacy before surgery and modify preoperative risk. We hypothesized that polypharmacy would be adversely associated with postoperative outcomes and healthcare resource use in a population older than 65 yr old who underwent elective, noncardiac surgery. We also sought to explore the possible contribution of confounding bias to this association.
Materials and Methods
Study Setting and Data Sources
After approval by the Sunnybrook Health Sciences Centre Research Ethics Board (Toronto, Canada), we conducted a population-based cohort study in Ontario, Canada, where hospital and physician services are provided to all residents through a publicly funded healthcare system and recorded in health administrative data sets that are collected using standardized methods.13,14  In Ontario, residents age 65 yr or older also receive universal pharamacare insurance benefits. All data were linked deterministically using encrypted patient-specific identifiers at the Institute for Clinical Evaluative Sciences (Toronto, Canada), an independent research institute that houses the health administrative data for the province of Ontario. Data sets used for the study included: the Ontario Drug Benefits Database, which captures prescription drug claims for residents age 65 yr and older; the Discharge Abstract Database, which captures all hospitalizations; the Ontario Health Insurance Plan database, which captures physician service claims; the Assistive Devices Program Database, which records receipt of medical devices; the Continuing Care Reporting System, which records details of long-term and respite care; the Home Care Database, which records details of home-based healthcare services received; the National Ambulatory Care Reporting System, which captures details of all emergency and outpatient care; and the Registered Persons Database, which captures all death dates for residents of Ontario. The analytic data set was created and processed by a trained data analyst using data normally collected at the Institute for Clinical Evaluative Sciences. Analysis was performed by a study analyst (C.A.W.) and the lead author (D.I.M.), and was overseen by the senior author (C.v.W.). The study protocol was registered at Clinicaltrials.gov (NCT03133182). This report is compliant with both strengthening the reporting of observational studies in epidemiology (STROBE) and reporting of studies conducted using observational routinely collected health data (RECORD) guidance for observational studies using routinely collected health data.15,16 
Study Population
We identified all patients 66 yr and older having one of the following elective, intermediate- to high-risk noncardiac surgeries: peripheral arterial bypass, carotid endarterectomy, open abdominal aortic aneurysm repair, endovascular abdominal aortic aneurysm repair, total hip replacement, total knee replacement, large bowel surgery, partial liver resection, pancreaticoduodenectomy, gastrectomy, esophagectomy, pneumonectomy, lobectomy, nephrectomy, or cystectomy. These are all sex-neutral, intermediate- to high-risk operations and have been used together to study outcomes for surgical patients in Ontario (see table of codes used in Supplemental Digital Content A, http://links.lww.com/ALN/B629).17–22  All admissions were elective, and the validity and reliability of codes used to identify these elective procedures have been confirmed through reabstraction.23,24  Surgeries were identified between April 1, 2002 (to coincide with the introduction of International Classification of Diseases, Tenth revision [ICD-10; to identify diagnoses] and Canadian Classification of Intervention [to identify procedures]), and March 31, 2014 (the latest time at which all data sets were complete when we conducted the study). This was a patient-level cohort, where only the first surgery for each participant was included.
Exposure
Our exposure of interest was the presence of polypharmacy in the 90 days before surgery. For each patient, we identified each unique drug for which a prescription was filled for that patient from the Ontario Drug Benefits Database in the 90 days before their surgery date through identification of each unique subclass of drugs that an individual was taking. We could not identify vitamins, supplements, or over-the-counter medications, so our drug exposure was limited to prescription agents.
The specific definition of polypharmacy varies in the literature between four or more and six or more unique drugs.8  For our primary exposure, we defined polypharmacy to be present if five or more prescriptions for unique drugs were filled; this was represented as a binary variable. We also coded lookback periods of 30 and 180 days before surgery to support sensitivity analyses.
Outcomes
The primary outcome was overall survival in the 90 days after surgery, which was determined from the Registered Persons Database (the gold standard source for mortality data in Ontario). In-hospital complications were identified using clusters of ICD-10 type 2 diagnostic codes (captured from the Discharge Abstract Database).25  Length of stay was calculated from the Discharge Abstract Database as the number of days from surgery to discharge. Institutional discharge was defined as being discharged to a nonhome location after the index hospitalization. Readmissions in the 30 days after surgery were identified as creation of a new record in the Discharge Abstract Database after hospital discharge. Institutional discharge and readmission outcomes were assessed only within the group of patients discharged alive from hospital. Total healthcare costs in the 90 days after surgery were calculated using standard methods to derive patient-level healthcare costs from the perspective of the provincial health insurance plan; costs were standardized to 2014 Canadian dollars.26 
Covariates
Demographics were identified from the Registered Persons Database. Standard methods were used to identify all Elixhauser comorbidities based on ICD-9 and ICD-10 codes from the Discharge Abstract Database in the 3 yr preceding surgery.27  The American Society of Anesthesiologists score for each patient was identified from physician billing. Frailty-defining diagnoses were identified using the Johns Hopkins Adjusted Clinical Groups frailty-defining diagnoses indicator.22,28,29  The Hospital-patient One-year Mortality Risk score was calculated for each patient. This score is an externally validated risk adjustment model with excellent discrimination (c-statistic 0.89 to 0.92) and calibration for predicting mortality risk in hospitalized patients.30  Higher Hospital-patient One-year Mortality Risk scores signify higher risk of death. We also recorded the year of surgery, any emergency department visits, or acute care hospitalizations in the year before surgery, whether a person lived in a rural or urban setting, and each participant’s neighborhood income quintile as a marker of socioeconomic status.
Sample Size
We included all individuals enrolled in our universal healthcare system having a qualifying surgery who met inclusion criteria; no a priori sample size calculation was performed. Based on an expected rate of 90-day mortality of 2%, we had greater than 99% power to detect a 50% relative difference in mortality between arms.
Analysis
SAS Enterprise Guide 6.1 (SAS Institute, USA) was used for all analyses. Patient characteristics were compared between exposure levels using absolute standardized differences. Standardized differences greater than 0.1 are suggested to represent a substantial difference.31 
Our primary outcome was analyzed using proportional hazards regression; adherence to the proportional hazards assumption was verified using log–negative–log plots. We performed unadjusted analysis and multilevel multivariable adjusted analysis, which clustered patients within hospitals using a robust sandwich covariance matrix estimate. We adjusted for sex (binary), age as a quadratic term (per a fractional polynomial analysis32 ), income quintile (five-level categorical variable), rural status (binary), American Society of Anesthesiologists score (categorically expressed as less than III, III, IV, or V), presence of a frailty-defining diagnosis (binary), emergency department visits in the previous year (categorically expressed as 0, 1, or more than 1), any acute hospitalization in the previous year (binary), Hospital-patient One-year Mortality Risk score (linear), each Elixhauser comorbidity (binary), year of surgery (as a restricted cubic spline with three knots), and surgery type (12-level categorical variable).
Secondary outcomes were also analyzed on an unadjusted and adjusted basis. Adjusted secondary analyses included the same covariates as the primary adjusted analysis. Generalized linear mixed models (PROC GLIMMIX) with binary response distributions and a log link were used to analyze complication rates, discharge disposition, and readmission rates, whereas costs were analyzed using a γ response distribution and log link (which is recommended for analysis of surgical cost).33  Clustering within hospitals was accounted for using a random intercept term. Postoperative length of stay was modeled as time to discharge using proportional hazards regression; clustering of patients in hospitals was achieved in the same manner as in the primary analysis. Because in-hospital death was a competing risk to hospital discharge, we calculated the subdistributional hazard ratio using the methods of Fine and Gray.34  In this analysis, hazard ratios more than 1 signify shorter length of stay.
Sensitivity Analyses
We performed several prespecified sensitivity analyses to test the robustness of our primary analysis. To determine whether the lookback (90 days) to identify prescription drugs impacted the prevalence of polypharmacy or its estimated measure of association with survival, we repeated the primary analysis using a 180-day lookback to identify unique prescription drugs. Because using polypharmacy as a binary exposure could contribute to information loss through categorization of a continuous variable, we also tested the number of drugs taken as a fractional polynomial.35  This analysis identified the linear form as the best continuous representation. Therefore, for our second sensitivity analysis, we included a linear term representing the total number of prescription drugs taken in place of the polypharmacy term in our primary adjusted model.
Polypharmacy is more prevalent and is associated with increased rates of adverse events in frail and multimorbid individuals36 ; therefore, if a causal relationship existed between polypharmacy and adverse outcomes, we would expect a greater effect size in people with frailty or multimorbidity. To evaluate this possibility, we prespecified and tested an interaction term between polypharmacy and the presence of a frailty-defining diagnosis. Post hoc, we recognized that further exploration of a confounded association versus true causation was needed. Therefore, we ran several more sensitivity analyses. First, we reran our primary analysis but replaced the Elixhauser comorbidities with the Johns Hopkins Aggregated Diagnosis Groups37  to test the robustness of our findings across comorbidity scoring systems. The Aggregated Diagnosis Groups uses administrative data from both inpatient and outpatient sources to identify diagnostic groups relevant to each patient. Each Aggregated Diagnosis Group was then weighted using the methods of Austin et al.,38  an approach that results in a high level of discrimination (c-statistic 0.917) and excellent calibration for predicting mortality. We also created an indicator variable to summate the number of Elixhauser comorbidities present and categorized patients based on the number of individual comorbidities present (0, 1 or 2, and 3 or more) and tested an interaction term between polypharmacy and this three-level categorical variable, as well as calculating the adjusted polypharmacy-survival association within each category. We also tested an interaction term between polypharmacy and surgery type and reran our primary analysis stratified by surgery type (as opposed to having it as a fixed effect) to evaluate whether differences in the baseline hazard between procedures might also bias results.
Peer reviewers suggested that further analyses be performed. First, we reran the primary analysis with a polypharmacy lookback window limited to 30 days before surgery. Next, we accounted for the risk profile of specific prescribed medications. We identified prescriptions for centrally acting medications (benzodiazepines,39  antipsychotics,39  anticonvulsants, or antidepressants40 ) and opioids41  (full list of medications in Supplemental Digital Content B, http://links.lww.com/ALN/B629). We then categorized participants by the number of high-risk medications that they were taking (0 was reference, 1, 2, or 3 or more) and used this variable in the primary adjusted model in place of the binary polypharmacy exposure. We also tested interaction terms by multiplying this high-risk medication count variable with frailty and with the Elixhauser comorbidity count variable.
Missing Data
No exposure, primary, or secondary outcome variables were missing. Neighborhood income quintile was missing for 0.4% of cases; the median value (third quintile) was imputed for these cases. Rural residency status was missing for 0.1% of cases, and in these cases the most common value (not rural) was imputed. No other data were missing.
Results
We identified 266,499 people age 66 yr or older who had intermediate- to high-risk elective noncardiac surgery from 2002 to 2014. Polypharmacy was present in 146,029 (54.8%) of individuals. Over the study period, the mean number of prescription medications taken increased significantly from 4.98 (SD 3.25) in 2002 to 2003, to 5.43 (SD 3.56) in 2013 to 2014 (P < 0.0001 for linear trend). Patients with polypharmacy were slightly older, were more likely to be female, and had a higher Hospital-patient One-year Mortality Risk score (table 1). The 20 most common classes of drugs prescribed for people with and without polypharmacy are shown in figure 1.
Table 1.
Baseline Characteristics of Study Population
Baseline Characteristics of Study Population×
Baseline Characteristics of Study Population
Table 1.
Baseline Characteristics of Study Population
Baseline Characteristics of Study Population×
×
Fig. 1.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
Fig. 1.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
×
Within 90 days of surgery, 6,275 patients died in the entire cohort (24 deaths per 1,000 patients). The crude mortality rate was higher in the polypharmacy group (4,356 deaths, 30 per 1,000 patients) compared to the group without polypharmacy (1,919 deaths, 16 per 1,000 patients; unadjusted hazard ratio 1.88; 95% CI, 1.78 to 1.99; P < 0.0001). After multilevel multivariable adjustment, polypharmacy continued to be significantly associated with decreased 90-day survival (adjusted hazard ratio 1.21; 95% CI, 1.14 to 1.27; P < 0.0001). The adjusted model parameters are provided in table 2. The area under the receiver operating curve for the model was 0.87, Nagelkerke’s R2 was 0.23, and the Akaike Information Criterions for the null and full models were 156,640 and 144,535, respectively.
Table 2.
Specification of Adjusted Survival Model (Clustered by Hospital)
Specification of Adjusted Survival Model (Clustered by Hospital)×
Specification of Adjusted Survival Model (Clustered by Hospital)
Table 2.
Specification of Adjusted Survival Model (Clustered by Hospital)
Specification of Adjusted Survival Model (Clustered by Hospital)×
×
Sensitivity Analyses for Survival
When polypharmacy was defined using a 180-day lookback, the adjusted association of polypharmacy with 90-day survival was essentially unchanged (adjusted hazard ratio 1.23; 95% CI, 1.14 to 1.32; P < 0.0001); however, limited to a 30-day lookback, the effect was attenuated (adjusted hazard ratio 1.07; 95% CI, 1.01 to 1.14; P = 0.02). When we tested the total number of prescription drugs as a continuous linear predictor instead of the binary polypharmacy term in our adjusted model, each additional drug was associated with a 1% decrease in 90-day survival (adjusted hazard ratio 1.01; 95% CI, 1.01 to 1.01; P < 0.0001). After adjustment for baseline health status with the Aggregated Diagnosis Groups Score (in place of Elixhauser comorbidities), polypharmacy continued to be significantly associated with decreased survival (adjusted hazard ratio 1.12; 95% CI, 1.05 to 1.20; P = 0.001).
The interaction between polypharmacy and frailty-defining diagnoses (P < 0.0001) was significant. Individuals with a frailty-defining diagnosis did not have a significant decrease in survival when exposed to polypharmacy (adjusted hazard ratio 0.92; 95% CI, 0.79 to 1.08), whereas the influence of polypharmacy was exaggerated in people without frailty-defining diagnoses compared to the full study cohort (adjusted hazard ratio 1.30; 95% CI, 1.21 to 1.41). The sum of Elixhauser comorbidities also interacted significantly with polypharmacy (P = 0.0002), and again polypharmacy had a decreased impact on survival where baseline health status was worse (adjusted hazard ratio with at least three comorbidities 0.96; 95% CI, 0.82 to 1.12; adjusted hazard ratio with 1 to 2 comorbidities 1.14; 95% CI, 1.02 to 1.26; adjusted hazard ratio with 0 comorbidities 1.30; 95% CI, 1.15 to 1.47). There was no statistically significant interaction between surgery type and polypharmacy (P = 0.18), and the procedure-stratified analysis was virtually unchanged from the primary analysis (adjusted hazard ratio 1.20; 95% CI, 1.14 to 1.27; P < 0.0001).
When we tested the count of centrally acting potentially high-risk drugs, each increase in the count was associated with decreased survival (0 was reference; 1 indicates adjusted hazard ratio 1.08; 95% CI, 1.02 to 1.14; 2 indicates adjusted hazard ratio 1.23; 95% CI, 1.14 to 1.32; 3 or more indicates adjusted hazard ratio 1.21; 95% CI, 1.09 to 1.35). There were significant associations between this high-risk medication variable and frailty (P < 0.0001) and Elixhauser index (P = 0.04). However, in people with frailty-defining diagnoses, there was no association between high-risk medications and survival, nor was there an association in the group of people with the highest comorbidity count (see results of all sensitivity analyses in Supplemental Digital Content C, http://links.lww.com/ALN/B629).
Secondary Outcomes
Rates, unadjusted, and adjusted measures of association for secondary outcomes are provided in table 3. Before adjustment, polypharmacy was significantly associated with all secondary outcomes. After multilevel multivariable adjustment, polypharmacy continued to be significantly associated with all secondary outcomes.
Table 3.
Secondary Study Outcomes
Secondary Study Outcomes×
Secondary Study Outcomes
Table 3.
Secondary Study Outcomes
Secondary Study Outcomes×
×
Discussion
In this population-based cohort study of older patients having major elective noncardiac surgery, patients who were exposed to polypharmacy in the 90 days before surgery had decreased postoperative survival, increased rates of complications, and higher resource use. This finding was independent of measured comorbidities, demographics, surgical risk factors, and baseline mortality risk. Therefore, the presence of polypharmacy identifies a high-risk stratum of the older surgical population. However, based on in-depth analysis of the interaction of preoperative medication counts and risk, with baseline health status, the potential impact of this association appears to be small or possibly nonexistent. Determining whether these associations represent true causation will require detailed prospective study that considers medication appropriateness and accounts for granular, patient-level variables such as functional and cognitive status.
Our study’s primary findings are consistent with the existing literature. To date, the majority of studies assessing the polypharmacy–outcome association have found that people with polypharmacy experience increased rates of adverse outcomes. In a systematic review of 50 studies involving community-dwelling older adults, consistent associations were found between polypharmacy and outcomes such as falls, adverse drug events, hospitalizations, and mortality.8  The effect sizes for mortality were similar to our primary findings, with hazard ratios ranging from 1.04 to 1.27. In the perioperative setting, the two studies that we were able to identify both found significant and adverse associations between preoperative medication burden and outcomes.9,10 
Several mechanisms could plausibly support the causal nature and generalizability of this association in perioperative patients. Polypharmacy could contribute to adverse drug–drug interactions, especially in the perioperative setting where new anesthetic and analgesic medication exposures will be present for all people and where pharmacokinetics can be altered by acute changes in end organ function. Furthermore, acute hospitalization is also associated with increased risk of postdischarge medication-related adverse events, which could contribute to poor postoperative outcomes.42,43  However, before claiming a causal association and recommending associated actions, the robustness of this association should be considered based on the current literature and our study’s findings.
Appraisal of the polypharmacy literature suggests that inadequate control for confounding and indication likely biases the results of most studies. In the systematic review by Fried et al.,8  studies were rated based on adjustment for comorbidity burden. In studies receiving the top rating of “good,” most studies found a significant association between polypharmacy and adverse outcomes. However, to receive a rating of “good,” studies were only required to adjust for a comorbidity score or the presence of multiple comorbidities. Given the complex interplay between polypharmacy and the indications for polypharmacy (such as comorbidity burden, overall health status, etc.), simple adjustment for comorbidity burden is inadequate for making causal inferences. In the limited evidence base assessing the association between polypharmacy and outcomes in the perioperative literature, control for confounding is also inadequate, with one study adjusting for age, duration of surgery, and use of cardiac drugs,9  and the second adjusting only for dementia, age, and sex.10 
Based on the proposed mechanisms underlying the polypharmacy–outcome association, one would expect that the risk of polypharmacy would be exaggerated in people with worse baseline health status. However, despite robust control for confounding (including a full set of Elixhauser comorbidities included as individual terms [as opposed to a prescored index which can bias results44 ], along with demographic factors, frailty, baseline mortality risk, preoperative health system utilization, and clustering to account for unmeasured hospital-level confounding), our models that assessed for effect modification by disease status found that strength of association between polypharmacy and survival was inversely related to baseline levels of illness. Furthermore, in a sensitivity analysis limited to higher-risk, centrally acting drugs, a similar pattern was found, with smaller effect sizes in groups with higher comorbidity burden and no association in people with frailty. A recently published study of polypharmacy and noncancer mortality found that increasingly robust control for confounding bias leads to decreasing strength of association between polypharmacy and mortality and a nonsignificant association of polypharmacy with survival in people with multimorbidity.45  Therefore, caution is warranted before altering perioperative practice based on the presence of polypharmacy.
If the polypharmacy-survival association is largely explained by confounding bias, true baseline illness status would have to be differentially misclassified. In this case, polypharmacy would be acting as a proxy for unmeasured illness in patients without documented comorbidities or frailty-defining diagnoses. When our analysis isolated the effect of polypharmacy to people with documented comorbidities, controlling for these disease states attenuated or removed the perceived effect of polypharmacy on survival. In analyses isolated to people without documented comorbidities, the presence of medications may have acted as a surrogate for undocumented illnesses that both lead to receipt of medications and may be the true causes of decreased survival. This is certainly plausible, because misclassification bias is a well-known limitation in studies using administrative data.46  Furthermore, the effect of polypharmacy on survival was decreased (hazard ratio 1.12 vs. 1.21), when the Aggregated Diagnosis Groups score (which sources diagnoses from inpatient and outpatient settings) replaced the Elixhauser Index (which sources diagnoses only from inpatient data) in our adjusted model.
If a causal relationship does exist, baseline illness would need to be a true effect modifier. In other words, older surgical patients with frailty or multimorbidity would have to benefit from exposure to multiple drugs. Perhaps medically complex patients are preferentially selected as appropriate candidates for elective surgery when their medical conditions are well managed using multiple appropriate agents. For patients with polypharmacy who are not frail or multimorbid, perhaps the impact of multiple or inappropriate agents, for which harm outweighs benefit, are amplified by the stress of the perioperative period. However, low risk of bias prospective study with highly accurate adjustment for comorbidity, frailty, function, cognition, and exposure status based on medication appropriateness (such as Beers47  or Screening Tool of Older People’s Prescriptions and Screening Tool to Alert to Right Treatment48  criteria) will be required to definitively answer this important question of causality in older surgical patients.
Strengths and Limitations
This study’s findings must be considered within the context of its strengths and limitations. As an observational study, we are unable to fully account for unmeasured confounders and indication bias. Despite adjustment for an extensive set of comorbidities identified through a 3-yr lookback period, preoperative health system use, a highly discriminative mortality risk score, and a frailty definition shown to predict death and increased resource use after surgery,22  we were not able to account for granular patient-level data such as physiologic or functional measures. Our study used health administrative data that were not initially collected for research purposes; therefore, one must also consider the risk of misclassification bias. Our exposure was based on health insurance data, which have a high degree of accuracy; however, it measures medications dispensed, not necessarily taken. Our outcomes were defined using the gold-standard source for mortality data as well as validated secondary outcomes measures. Our polypharmacy exposure was a dichotomization of a continuous variable and did not account for the potential differing risk impacts or appropriateness of different drugs. We were also unable to account for medication underutilization (i.e., patients not being dispensed medications that would be expected to improve their health outcomes), which could bias our results toward the null.
Conclusions
Polypharmacy is present in the majority of older patients having elective major noncardiac surgery and is associated with decreased postoperative survival, increased adverse event rates, and higher health resource utilization. However, given the decreased effect sizes observed in subgroups of sicker patients, our results call into question the interpretation of causality in the polypharmacy–outcome association. Low risk of bias prospective studies are needed to estimate the true risk of polypharmacy in older surgical patients.
Research Support
Dr. McIsaac receives salary support from the Department of Anesthesiology, Ottawa Hospital (Ottawa, Ontario, Canada) and the Canadian Anesthesiology Society’s Career Scientist Award. This study was also supported by the Institute for Clinical Evaluative Sciences (Toronto, Canada), which is funded by an annual grant from the Ontario Ministry of Health and Long-term Care (Toronto, Canada). The opinions, results, and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by Institute for Clinical Evaluative Sciences or the Ontario Ministry of Health and Long-term Care is intended or should be inferred. These data sets were held securely in a linked, deidentified form, and were analyzed at the Institute for Clinical Evaluative Sciences.
Competing Interests
The authors declare no competing interests.
References
Etzioni, DA, Liu, JH, Maggard, MA, Ko, CY The aging population and its impact on the surgery workforce. Ann Surg 2003; 238:170–7 [PubMed]
Turrentine, FE, Wang, H, Simpson, VB, Jones, RS Surgical risk factors, morbidity, and mortality in elderly patients. J Am Coll Surg 2006; 203:865–77 [Article] [PubMed]
Hamel, MB, Henderson, WG, Khuri, SF, Daley, J Surgical outcomes for patients aged 80 and older: Morbidity and mortality from major noncardiac surgery. J Am Geriatr Soc 2005; 53:424–9 [Article] [PubMed]
The Anaesthesia and Perioperative Care Priority Setting Partnership 2015. Available at: http://niaa.org.uk/Results#pt. Accessed June 8, 2017
Arnoldo, L, Cattani, G, Cojutti, P, Pea, F, Brusaferro, S Monitoring polypharmacy in healthcare systems through a multi-setting survey: Should we put more attention on long term care facilities? J Public Health Res 2016; 5:745 [Article] [PubMed]
Gnjidic, D, Hilmer, SN, Blyth, FM, Naganathan, V, Waite, L, Seibel, MJ, McLachlan, AJ, Cumming, RG, Handelsman, DJ, Le Couteur, DG Polypharmacy cutoff and outcomes: Five or more medicines were used to identify community-dwelling older men at risk of different adverse outcomes. J Clin Epidemiol 2012; 65:989–95 [Article] [PubMed]
Guthrie, B, Makubate, B, Hernandez-Santiago, V, Dreischulte, T The rising tide of polypharmacy and drug-drug interactions: Population database analysis 1995-2010. BMC Med 2015; 13:74 [Article] [PubMed]
Fried, TR, O’Leary, J, Towle, V, Goldstein, MK, Trentalange, M, Martin, DK Health outcomes associated with polypharmacy in community-dwelling older adults: A systematic review. J Am Geriatr Soc 2014; 62:2261–72 [Article] [PubMed]
Kennedy, JM, van Rij, AM, Spears, GF, Pettigrew, RA, Tucker, IG Polypharmacy in a general surgical unit and consequences of drug withdrawal. Br J Clin Pharmacol 2000; 49:353–62 [Article] [PubMed]
Jeong, YM, Lee, E, Kim, KI, Chung, JE, In Park, H, Lee, BK, Gwak, HS Association of pre-operative medication use with post-operative delirium in surgical oncology patients receiving comprehensive geriatric assessment. BMC Geriatr 2016; 16:134 [Article] [PubMed]
Härstedt, M, Rogmark, C, Sutton, R, Melander, O, Fedorowski, A Polypharmacy and adverse outcomes after hip fracture surgery. J Orthop Surg Res 2016; 11:151 [Article] [PubMed]
Munson, JC, Bynum, JP, Bell, JE, Cantu, R, McDonough, C, Wang, Q, Tosteson, TD, Tosteson, AN Patterns of prescription drug use before and after fragility fracture. JAMA Intern Med 2016; 176:1531–8 [Article] [PubMed]
Vision and Mandate. Available at: https://www.cihi.ca/en/submit-data-view-standards/data-standards. Accessed January 30, 2017
About Us, Institute for Clinical Evaluative Sciences website. Available at: http://www.ices.on.ca/webpage.cfm?site_id=1&org_id=26. Accessed June 8, 2017
Elm, E von, Altman, DG, Egger, M, Pocock, SJ, Gøtzsche, PC, Vandenbroucke, JP Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. BMJ 2007; 335:806–8 [Article] [PubMed]
Benchimol, EI, Smeeth, L, Guttmann, A, Harron, K, Moher, D, Petersen, I, Sørensen, HT, von Elm, E, Langan, SM RECORD Working Committee: The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med 2015; 12:e1001885 [Article] [PubMed]
Wijeysundera, DN, Austin, PC, Beattie, WS, Hux, JE, Laupacis, A Variation in the practice of preoperative medical consultation for major elective noncardiac surgery: A population-based study. Anesthesiology 2012; 116:25–34 [Article] [PubMed]
Wijeysundera, DN, Wijeysundera, HC, Yun, L, Wąsowicz, M, Beattie, WS, Velianou, JL, Ko, DT Risk of elective major noncardiac surgery after coronary stent insertion: A population-based study. Circulation 2012; 126:1355–62 [Article] [PubMed]
Wijeysundera, DN, Beattie, WS, Karkouti, K, Neuman, MD, Austin, PC, Laupacis, A Association of echocardiography before major elective non-cardiac surgery with postoperative survival and length of hospital stay: Population based cohort study. BMJ 2011; 342:d3695 [Article] [PubMed]
Wijeysundera, DN, Austin, PC, Beattie, WS, Hux, JE, Laupacis, A A population-based study of anesthesia consultation before major noncardiac surgery. Arch Intern Med 2009; 169:595–602 [Article] [PubMed]
McIsaac, DI, Bryson, GL, van Walraven, C Elective, major noncardiac surgery on the weekend: A population-based cohort study of 30-day mortality. Med Care 2014; 52:557–64 [Article] [PubMed]
McIsaac, DI, Bryson, GL, van Walraven, C Association of frailty and 1-year postoperative mortality following major elective noncardiac surgery: A population-based cohort study. JAMA Surg 2016; 151:538–45 [Article] [PubMed]
Bourne, R, DeBoer, D, Hawker, G, Kreder, H, Mahomed, M, Paterson, J Tu, J V, Laupacis, A, Pinfold, S, McColgan, P Total hip and knee replacement, Access to Health Service in Ontario: ICES Atlas. 2005 Toronto, Institute for Clinical and Evaluative Sciences,
Technical supplement: Health Care in Canada 2005. 2005. Available at: https://secure.cihi.ca/estore/productSeries.htm?pc=PCC64. Accessed June 8, 2017
Southern, DA, Burnand, B, Droesler, SE, Flemons, W, Forster, AJ, Gurevich, Y, Harrison, J, Quan, H, Pincus, HA, Romano, PS, Sundararajan, V, Kostanjsek, N, Ghali, WA Deriving ICD-10 codes for patient safety indicators for large-scale surveillance using administrative hospital data. Med Care 2017; 55:252–60 [Article] [PubMed]
Wodchis, W, Bushmeneva, K, Nikitovic, M, McKillop, IGuidelines on person-level costing using administrative databases in Ontario. 2013. Toronto, Available at: http://www.hsprn.ca/uploads/files/Guidelines_on_PersonLevel_Costing_May_2013.pdf. Accessed May 1, 2017
Quan, H, Sundararajan, V, Halfon, P, Fong, A, Burnand, B, Luthi, JC, Saunders, LD, Beck, CA, Feasby, TE, Ghali, WA Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43:1130–9 [Article] [PubMed]
Lieberman, R, Abrams, C, Weiner, J Development and evaluation of the Johns Hopkins University risk adjustment models for Medicare+Choice plan payment. 2003 Baltimore,
Sternberg, SA, Bentur, N, Abrams, C, Spalter, T, Karpati, T, Lemberger, J, Heymann, AD Identifying frail older people using predictive modeling. Am J Manag Care 2012; 18:e392–7 [PubMed]
Walraven, C van, McAlister, FA, Bakal, JA, Hawken, S, Donze, J External validation of the Hospital-patient One-year Mortality Risk (HOMR) model for predicting death within 1 year after hospital admission. C Can Med Assoc J 2015; 187:725–33 [Article]
Austin, PC Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat - Simul Comput 2009; 38:1228–34 [Article]
Sauerbrei, W, Meier-Hirmer, C, Benner, A, Royston, P Multivariable regression model building by using fractional polynomials: Description of SAS, STATA and R programs. Comput Stat Data Anal 2006; 50:3464–85 [Article]
Austin, PC, Ghali, WA, Tu, JV A comparison of several regression models for analysing cost of CABG surgery. Stat Med 2003; 22:2799–815 [Article] [PubMed]
Fine, JP, Gray, RJ A Proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94:496 [Article]
Sauerbrei, W, Royston, P Building multivariable prognostic and diagnostic models: Transformation of the predictors by using fractional polynomials. J R Stat Soc Ser A Stat Soc 2002; 165:399–400 [Article]
Poudel, A, Peel, NM, Nissen, LM, Mitchell, CA, Gray, LC, Hubbard, RE Adverse outcomes in relation to polypharmacy in robust and frail older hospital patients. J Am Med Dir Assoc 2016; 17:767.e9–767.e13 [Article]
Technical Reference Guide 9.0. 2009. Baltimore, Available at: http://acg.jhsph.org. Accessed January 30, 2017
Austin, PC, van Walraven, C, Wodchis, WP, Newman, A, Anderson, GM Using the Johns Hopkins Aggregated Diagnosis Groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada. Med Care 2011; 49:932–9 [Article] [PubMed]
Maclagan, LC, Maxwell, CJ, Gandhi, S, Guan, J, Bell, CM, Hogan, DB, Daneman, N, Gill, SS, Morris, AM, Jeffs, L, Campitelli, MA, Seitz, DP, Bronskill, SE Frailty and potentially inappropriate medication use at nursing home transition. J Am Geriatr Soc 2017; 65:2205–12 [Article] [PubMed]
Brown, HK, Ray, JG, Wilton, AS, Lunsky, Y, Gomes, T, Vigod, SN Association between serotonergic antidepressant use during pregnancy and autism spectrum disorder in children. JAMA 2017; 317:1544–52 [Article] [PubMed]
Clarke, H, Soneji, N, Ko, DT, Yun, L, Wijeysundera, DN Rates and risk factors for prolonged opioid use after major surgery: Population based cohort study. BMJ 2014; 348:g1251 [Article] [PubMed]
Moore, C, Wisnivesky, J, Williams, S, McGinn, T Medical errors related to discontinuity of care from an inpatient to an outpatient setting. J Gen Intern Med 2003; 18:646–51 [Article] [PubMed]
Forster, AJ, Murff, HJ, Peterson, JF, Gandhi, TK, Bates, DW The incidence and severity of adverse events affecting patients after discharge from the hospital. Ann Intern Med 2003; 138:161–7 [Article] [PubMed]
Schneeweiss, S, Maclure, M Use of comorbidity scores for control of confounding in studies using administrative databases. Int J Epidemiol 2000; 29:891–8 [Article] [PubMed]
Schöttker, B, Saum, K-U, Muhlack, DC, Hoppe, LK, Holleczek, B, Brenner, H Polypharmacy and mortality: New insights from a large cohort of older adults by detection of effect modification by multi-morbidity and comprehensive correction of confounding by indication. Eur J Clin Pharmacol 2017:1–8
Walraven, C van, Austin, P Administrative database research has unique characteristics that can risk biased results. J Clin Epidemiol 2012; 65:126–31 [Article] [PubMed]
American Geriatrics Society 2015 Beers Criteria Update Expert Panel: American Geriatrics Society 2015 Updated Beers Criteria for potentially inappropriate medication use in older adults. J Am Geriatr Soc 2015; 63:2227–46 [Article] [PubMed]
O’Mahony, D, O’Sullivan, D, Byrne, S, O’Connor, MN, Ryan, C, Gallagher, P STOPP/START criteria for potentially inappropriate prescribing in older people: Version 2. Age Ageing 2015; 44:213–8 [Article] [PubMed]
Fig. 1.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
Fig. 1.
The 20 most common drug classes dispensed in the 90 days before surgery for people taking less than five drugs concurrently (A) and people taking five or more drugs concurrently (B). ACE = angiotensin-converting enzyme.
×
Table 1.
Baseline Characteristics of Study Population
Baseline Characteristics of Study Population×
Baseline Characteristics of Study Population
Table 1.
Baseline Characteristics of Study Population
Baseline Characteristics of Study Population×
×
Table 2.
Specification of Adjusted Survival Model (Clustered by Hospital)
Specification of Adjusted Survival Model (Clustered by Hospital)×
Specification of Adjusted Survival Model (Clustered by Hospital)
Table 2.
Specification of Adjusted Survival Model (Clustered by Hospital)
Specification of Adjusted Survival Model (Clustered by Hospital)×
×
Table 3.
Secondary Study Outcomes
Secondary Study Outcomes×
Secondary Study Outcomes
Table 3.
Secondary Study Outcomes
Secondary Study Outcomes×
×