Abstract
Background
Injection drug use–associated infective endocarditis (IDU-IE) is rising and valve surgery is frequently indicated. The effect of initiating public outcomes reporting for aortic valve surgery on rates of valve surgery and in-hospital mortality for endocarditis is not known.
Methods
For an interrupted time series analysis, we used data from the National Inpatient Sample, a representative sample of United States inpatient hospitalizations, from January 2010 to September 2015. We included individuals aged 18–65 with an International Classification of Diseases, Ninth Revision (ICD-9) diagnosis of endocarditis. We defined IDU-IE using a validated combination of ICD-9 codes. We used segmented logistic regression to assess for changes in valve replacement and in-hospital mortality rates after the public reporting initiation in January 2013.
Results
We identified 7322 hospitalizations for IDU-IE and 23 997 for non–IDU-IE in the sample, representing 36 452 national IDU-IE admissions and 119 316 non-IDU admissions, respectively. Following the implementation of public reporting in 2013, relative to baseline trends, the odds of valve replacement decreased by 4.0% per quarter (odds ratio [OR] 0.96, 95% confidence interval [CI] 0.93–0.99), with no difference by IDU status. The odds of an in-patient death decreased by 2.0% per quarter for both IDU-IE and non–IDU-IE cases following reporting (OR 0.98, 95% CI 0.97–0.99).
Conclusions
Initiating public reporting was associated with a significant decrease in valve surgery for all IE cases, regardless of IDU status, and a reduction in-hospital mortality for patients with IE. Patients with IE may have less access to surgery as a consequence of public reporting. To understand how reduced valve surgery impacts overall mortality, future studies should examine the postdischarge mortality rate.
Keywords: endocarditis, injection drug use, cardiac surgery, public reporting
In this interrupted time-series analysis of United States hospitalizations, the initiation of public reporting of aortic valve surgery outcomes in 2013 was associated with a 4.0% decrease in odds of surgery per quarter, with no differences by injection drug status.
(See the Editorial Commentary by Springer on pages 488–90.)
Injection drug–associated infective endocarditis (IDU-IE) hospitalizations doubled in the United States between 2000 and 2013 and, along with more recent dramatic regional increases, contribute to rising endocarditis rates [1–5]. Endocarditis is generally treated with prolonged intravenous antibiotics and frequently requires valve replacement [5, 6]. Valve replacement reduces the risk of mortality for patients with heart failure, an abscess, heart block, resistant organisms, and persistent bacteremia (Class I surgical indications), as well as recurrent emboli, severe regurgitation, or mobile vegetations >10 mm (Class II surgical indications) [6, 7]. Surgical rates increased by 7% per decade from 1969 to 2000 [8], with some studies reporting surgical rates of nearly 50% by 1999 [6, 9–11]. A 2012 trial that demonstrated reduced rates of death and embolism among those receiving early surgery bolstered this trend [12]. Nevertheless, surgeons may hesitate to replace valves in those with IDU-IE, out of concern for complications from drug use [6, 13–15], despite equivalent short-term surgical outcomes to non–IDU-IE patients [15, 16], even without addiction care [17–20]. This hesitancy may reflect negative attitudes towards people with IDU-IE (eg, overestimating the surgical risk) and could be compounded by attention to outcomes.
After collecting data for a year, in January 2013, the Society of Thoracic Surgery (STS) began publicly reporting risk-adjusted in-hospital and 30-day mortality and morbidity rates after aortic valve replacement (AVR), as part of an effort to sequentially expand reporting to several cardiac surgeries [21–23]. Of nearly 30 000 AVRs performed annually, most are for indications other than endocarditis, like aortic stenosis, which has an in-hospital mortality rate of <2% [22, 24, 25]. In comparison, the in-hospital mortality rate for endocarditis is higher (15% in some studies [5]), exposing surgeons to an increased risk of poor outcomes, despite STS risk adjustments for endocarditis and drug use [26, 27].
Numerous studies have demonstrated that public reporting promotes transparency and quality improvements, including in cardiac surgery [25, 28]. These measures also result in fewer indicated procedures, like percutaneous coronary intervention [29, 30] and coronary bypass graft surgery [31, 32] for high-risk patients. Our objective was to determine whether increased scrutiny on outcomes through public reporting for AVR affected the odds of receiving valve surgery or the in-hospital mortality rate among patients hospitalized with IDU-IE and non–IDU-IE, and whether changes differed by IDU status. We hypothesized that following public reporting, the odds of receiving surgery in all endocarditis would decline, but to a greater degree among those with IDU-IE, and that in-hospital mortality rates would increase among both IDU-IE and non–IDU-IE patients.
METHODS
Study Design and Data Source
We used an interrupted time series design (ITS) to investigate the effect of initiating public reporting of AVR outcomes in January 2013 on rates of valve surgery and in-hospital mortality in IDU and non–IDU-IE cases. We used data collected between 1 January 2010 and 31 August 2015 from the National Inpatient Sample (NIS). The NIS is a large, publicly available database comprising a 20% stratified sample of all inpatient hospitalizations in the United States, excluding hospitalizations at Veterans Affairs’ hospitals. The Boston University Medical Campus Institutional Review Board determined this analysis was not human subjects research.
Cohort Selection
We included hospitalizations among individuals aged 18 to 65 with a diagnosis of endocarditis not associated with a prosthetic valve, using International Classification of Diseases, Ninth Revision (ICD-9) codes assigned during the study period (421.0, 421.1, 421.9, 424.90, 424.91, 424.99, and 112.81). We selected these ages to reduce heterogeneity between the comparison groups and limit the identification of congenital heart disease or structural heart disease in older individuals, consistent with previous IDU-IE studies [1, 2, 5, 20]. We excluded hospitalizations resulting in transfers to another acute-care facility, to prevent double counting. Hospitalizations with missing data were excluded from the analyses: 35 946 hospitalizations for IE were identified. Hospitalizations with missing admission month (1304), race (2304), payor (110), zip code income quartile (834), and elective status (75) data were excluded. The study sample was comprised of the remaining 31 319 hospitalizations. Mortality data were missing for 42 hospitalizations, which were excluded from analyses where mortality was an outcome, resulting in a sample of 31 277 hospitalizations.
We stratified the cohort into those patients with IDU-IE and those with non–IDU-IE. We defined IDU-IE using ICD-9 diagnosis codes corresponding to injection drug use or hepatitis C, based on a previously validated algorithm found to have a sensitivity of 93% and specificity of 61%, with an 83% positive predictive value for identifying IE resulting from IDU [33]. We converted ICD-10 codes from this algorithm to ICD-9 codes using 2015 general equivalence mapping from the Centers for Medicaid and Medicare Services (Supplementary Table 1) [34].
Variables of Interest
The primary outcome was receipt of valve surgery during hospitalization for endocarditis, using ICD-9 procedure codes (Supplementary Table 2). While public reporting measures focused on AVR, we defined the outcome as any valve surgery, because not all diagnosis codes specify affected valve. Surgeons may have contextualized this policy as part of the broader effort to report outcomes (STS planned to report mitral valve outcomes in 2016). As a secondary outcome, we reported in-hospital mortality rates, as documented in the NIS.
We analyzed several covariates available in the NIS, including age, sex, race, insurance status, income quartile of the patient’s zip code, hospital region, and elective admission. We included illness severity and mortality risk scores, which classify the risk of disability and mortality, respectively, on 1 to 4 scales, with 4 representing an extreme loss of function or likelihood of death, respectively, as assigned by All Patient Refined Diagnosis Related Groups 3M software in the NIS [35, 36].
Statistical Analysis
We analyzed outcomes on a quarterly basis and divided the study into 2 periods: preintervention (Quarter 1 [Q1]/2010 to Q4/2012) and postintervention (Q1/2013 through August of Q3/2015). We plotted the unadjusted proportion of hospitalizations undergoing valve surgery and proportion with in-hospital mortality for each quarter. We used segmented logistic regression to test for changes in the odds of receiving valve surgery and the in-hospital mortality rates by quarter associated with the intervention. Regression models included terms for the baseline level and trend, beginning with the first postintervention quarter (Q1/2013) [37]. We included interaction terms between these 4 regression terms and IDU status, to identify differences in the trend and level of outcome by IDU status prior to the intervention and the differential effect of the intervention by IDU status. To improve the power to detect significant predictors and avoid nonsignificant terms, we used backward elimination to select the most parsimonious model, sequentially removing terms with P values > .20 [37]. We used covariates to adjust for potential confounders that varied with time. We used regression results to estimate absolute effects and 95% confidence intervals [CIs] for each outcome, stratified by IDU-IE and non–IDU-IE cases, using sample means and relevant proportions. We used survey procedures in SAS, version 9.4 (SAS Institute Inc., Cary, NC), to account for the NIS’s complex survey design.
Sensitivity Analyses
We performed multiple sensitivity analyses. First, to explore the impact of missingness, we included observations with missing data for any variable other than time, recoding the variable as missing. Second, we excluded elective admissions, where a plan for valve surgery may have already been determined. Third, as public reporting focused on AVR, we limited the surgical outcome definition to ICD-9 codes identifying aortic valve procedures (Supplementary Table 2). Fourth, to test whether any changes observed were centered at the time of the intervention, including whether there was a delayed effect, we repeated the final outcomes models, shifting the time of the trend and level change variables from 4 quarters before and after Q1/2013, and the compared goodness of fit using the Akaike Information Criterion. Fifth, we hypothesized that mortality benefits would accrue differentially in patients receiving surgery if higher-risk candidates were denied surgery and, thus, repeated the mortality analysis, stratified by receipt of surgery, and compared mean mortality and illness severity scores in these groups before and after public reporting. Sixth, we removed hepatitis C from the algorithm used to classify endocarditis as IDU-IE to explore the effect of any misclassifications based on this diagnosis. Seventh, we tested whether hospitalizations resulting in transfers changed over time and repeated analyses including hospitalizations resulting in transfers in the cohort.
RESULTS
We identified 31 319 IE hospitalizations in the sample, corresponding to 178 781 estimated national hospitalizations (Table 1). IDU-IE cases accounted for 23.4% (36 452) and non–IDU-IE cases for 76.6% (119 316) of all IE hospitalizations. The proportion of all IE cases classified as IDU-IE increased from 18.3% in the preintervention period to 27.3% in the postintervention period. Compared to non–IDU-IE patients, IDU-IE patients were younger, more likely to be White, more likely to be insured by Medicaid, came from the lowest zip code income quartile, and less likely to be admitted electively or in the Midwest. Of the IDU-IE hospitalizations, 12 392 were during the preintervention period and 24 060 were in the postintervention period. IDU-IE patients hospitalized in the postintervention period were younger, more likely to be female and White, more likely to have Medicaid, and less likely to live in the Northeast, compared to IDU-IE patients hospitalized in the preintervention period. Of the non–IDU-IE patients hospitalizated, 55 316 cases were in the preintervention period and 64 000 were in the postintervention period. In the postintervention period, individuals with non-IDU hospitalizations were less likely to be female or elective admissions, and more likely to be White and insured by Medicaid.
Table 1.
Characteristics of Infectious Endocarditis Hospitalizations
| IDU | No IDU | |||||
|---|---|---|---|---|---|---|
| Preintervention | Postintervention | Preintervention | Postintervention | |||
| n = 2510 | n = 4812 | n = 11 197 | n = 12 800 | |||
| Weighted, n = 12 392 | Weighted, n = 24 060 | Weighted, n = 55 316 | Weighted, n = 64 000 | |||
| (SE 533) | (SE 525) | P Value | (SE 1571) | (SE 938) | P Value | |
| Age, meanb (SE) | 41.2 (0.3) | 38.5 (0.2) | <.0001 | 49.7 (0.2) | 49.6 (0.1) | <.0001a |
| Female, % | 39.2 | 45.5 | <.0001 | 42.7 | 39.7 | <.0001a |
| Race, % | … | … | <.0001 | … | … | .023a |
| White | 65.0 | 76.0 | … | 60.8 | 62.6 | … |
| Black | 20.2 | 12.2 | … | 24.1 | 21.5 | … |
| Hispanic | 10.4 | 8.5 | … | 9.0 | 10.2 | … |
| Asian | 0.7 | 0.6 | … | 2.2 | 2.2 | … |
| Native American | 0.8 | 0.8 | … | 0.7 | 0.8 | … |
| Other | 2.9 | 2.1 | … | 3.2 | 2.8 | … |
| Payor, % | … | … | <.0001 | … | … | .009a |
| Medicare | 16.7 | 13.2 | … | 32.1 | 30.8 | … |
| Medicaid | 39.4 | 49.0 | … | 20.4 | 23.2 | … |
| Commercial | 13.8 | 12.8 | … | 35.3 | 35.1 | … |
| Self | 22.0 | 18.9 | … | 7.6 | 7.0 | … |
| No charge | 1.9 | 2.7 | … | 0.8 | 0.8 | … |
| Other | 6.2 | 3.5 | … | 3.8 | 3.3 | … |
| Region, % | … | … | <.0001 | … | … | .1a |
| Northeast | 23.9 | 18.0 | … | 21.1 | 18.5 | … |
| Midwest | 13.5 | 14.8 | … | 18.3 | 19.0 | … |
| South | 36.5 | 44.9 | … | 41.0 | 44.1 | … |
| West | 26.1 | 22.3 | … | 19.6 | 18.4 | … |
| Zip code income quartile, % | … | … | .0014 | … | … | .07a |
| 0–25th% | 42.0 | 42.2 | … | 35.9 | 36.2 | … |
| 26–50th% | 22.6 | 27.2 | … | 24.0 | 26.0 | … |
| 51st–75% | 20.8 | 18.4 | … | 22.0 | 20.9 | … |
| 76th–100th% | 14.6 | 12.2 | … | 20.9 | 16.9 | … |
| Elective, % | 8.0 | 6.9 | .16 | 14.7 | 12.0 | <.0001a |
| Length of stay, days (SE) | 15.9 (0.5) | 16.1 (0.3) | <.0001 | 13.7 (0.2) | 13.1 (0.2) | <.0001a |
| Mortality risk scorec (SE) | 2.9 (0.03) | 3.1 (0.02) | <.0001 | 2.9 (0.01) | 3.0 (0.009) | <.0001a |
| Illness severity scored (SE) | 3.4 (0.02) | 3.4 (0.01) | <.0001 | 3.3 (0.01) | 3.3 (0.007) | <.0001a |
Data are shown by IDU status and before and after public reporting intervention (1 January 2010–31 August 2015). P values from Chi-squared testing for categorical variables and t-tests for continuous variables were used to assess for within-group differences between preintervention versus postintervention periods. These data reflect the cardiac surgery cohort; an additional 42 hospitalizations were excluded for mortality analysis due to missing death data.Abbreviations: DRG, diagnosis related group; IDU, injection drug use; SE, standard error.
aAll values are weighted estimates.
bIndicates statistically significant differences between IDU and non-IDU cohorts for the specified variable.
cAll patient-refined DRG, Risk of Mortality on 1–4 Scale where 4 is extreme likelihood of dying.
dAll patient-refined DRG, Risk of Loss of Function on 1–4 Scale where 4 is extreme loss of function.
Valve Surgery Rates
In the preintervention period, 11.4% (95% CI 10.0%-12.9%) of IDU-IE and 12.0% (95% CI 11.1–13.0) of non–IDU-IE hospitalizations included valve surgery. In the postintervention period, the proportion of hospitalizations with valve surgery were 12.0% (95% CI 11.0%–13.0%) of IDU-IE and 13.6% (95% CI 12.9%–14.3%) of non–IDU-IE cases (Supplementary Figure 1). In adjusted segmented regression models after backwards selection (Supplementary Table 3), among non–IDU-IE patients, the odds of receiving valve surgery increased by 3% in each quarter during the preintervention period (adjusted odds ratio [AOR] 1.03, 95% CI 1.01–1.05; Table 2). Relative to non–IDU IE patients, the odds of receiving surgery decreased by 1.0% quarterly for IDU-IE patients (AOR 0.99, 95% CI 0.98–0.99). Following public reporting in Q1/2013, the odds of receiving surgery declined by 4% each quarter (AOR 0.96, 95% CI 0.93–0.99) for both IDU-IE and non–IDU-IE patients, without a differential change by IDU status. Comparing preintervention trends to postintervention observations among IDU-IE patients, 2 years after the implementation of public reporting in Q1/2015, the probability of receiving valve surgery for IDU-IE, adjusting for changing characteristics of the population, decreased from a projected rate of 13.0 % (95% CI 7.5–21.5) to an observed rate of 9.8% (95% CI 6.2–15.0), a 33% relative decline (Figure 1; Table 3). Similarly, non-IDU hospitalizations experienced a decline from the projected preintervention trend rate of 16.3% (95% CI 10.4–24.7) to an observed rate of 12.4% (95% CI 8.0–18.8) following public reporting, for a 31% relative decline.
Table 2.
Segmented Logistic Regression Model Results
| Valve Surgery | Inpatient Mortality | |
|---|---|---|
| Model specifications | ||
| Baseline trend | 1.03 (1.01–1.05) | … |
| IDU | … | 0.79 (0.71–0.89) |
| IDU x Baseline trend | 0.99 (0.98–0.99) | … |
| Trend change | 0.96 (0.93–0.99) | 0.98 (0.97–0.99) |
| Age, per quarter | 0.99 (0.98–0.99) | 1.03 (1.02–1.03) |
| Female | 0.65 (0.6–0.7) | 0.94 (0.87–0.99) |
| Race | ||
| White | 1 (Referent) | 1 (Referent) |
| Black | 0.93 (0.82–1.06) | 0.96 (0.84–1.09) |
| Hispanic | 0.94 (0.8–1.09) | 0.92 (0.79–1.07) |
| Asian | 0.91 (0.72–1.15) | 1.18 (0.91–1.53) |
| Native American | 0.95 (0.64–1.41) | 1.09 (0.74–1.62) |
| Other | 1.18 (0.96–1.45) | 1.04 (0.83–1.29) |
| Payor | ||
| Medicaid | 1 (Referent) | 1 (Referent) |
| Medicare | 0.64 (0.58–0.71) | 1.07 (0.96–1.19) |
| Commercial | 1.32 (1.2–1.45) | 0.88 (0.79–0.98) |
| Self | 0.97 (0.86–1.1) | 1.12 (0.98–1.29) |
| No charge | 1.2 (0.87–1.64) | 0.86 (0.62–1.18) |
| Other | 1.11 (0.93–1.33) | 1.05 (0.87–1.28) |
| Zip code income quartile | ||
| Quartile 1 | 0.89 (0.84–0.96) | 1.10 (1.03–1.17) |
| Quartile 2 | 0.97 (0.91–1.03) | 1.02 (0.95–1.10) |
| Quartile 3 | 1.04 (0.98–1.12) | 0.93 (0.86–1.01) |
| Quartile 4 | 1 (Referent) | 1 (Referent) |
| Hospital region | ||
| Northeast | 1 (Referent) | 1 (Referent) |
| Midwest | 1.02 (0.93–1.12) | 0.95 (0.87–1.04) |
| South | 0.95 (0.88–1.03) | 1.03 (0.97–1.10) |
| West | 0.90 (0.82–0.99) | 1.00 (0.92–1.09) |
| Mortality risk score (0–4) | ||
| 0 | 12.61 (2.21–72.08) | 0.33 (0.03–3.41) |
| 1 | 1 (Referent) | 1 (Referent) |
| 2 | 0.76 (0.49–1.2) | 0.46 (0.24–0.86) |
| 3 | 1.09 (0.71–1.68) | 1.62 (0.97–2.70) |
| 4 | 1.65 (1.08–2.52) | 13.03 (7.87–21.56) |
| Disability risk score (0–4) | ||
| 0 | 0.11 (0.06–0.18) | 9.7 (1.39–67.48) |
| 1 | 1 (Referent) | 1 (Referent) |
| 2 | 1.24 (1.03–1.49) | 0.35 (0.17–0.70) |
| 3 | 0.89 (0.8–1.0) | 0.51 (0.42–0.62) |
| 4 | 1 | 1 |
| Non-elective admission | 0.7 (0.66–0.73) | 1.01 (0.93–1.08) |
Data are for valve surgery and inpatient mortality following implementation of public reporting (1 January 2010–31 August 2015). Values are adjusted odds ratios. IDU x Trend Change, Level Change, and IDU x Level Change are excluded from the table as these terms were removed from presented models. These are odds ratios which require a “referent” to which the other variables are compared. Abbreviation: IDU, injection drug use.
Figure 1.
Estimated trends in valve surgery and inpatient mortality following the implementation of public reporting (1 January 2010–31 August 2015). Estimates were derived from segmented regression models, using sample means and relevant proportions for each covariate. Abbreviations: IDU, injection drug use; Q, quartile.
Table 3.
Projected Baseline Trend Versus Observed Proportion Undergoing Cardiac Surgery and With Mortality
| Outcome | Projected Q1 2015, % | Observed Q1 2015, % | Relative Difference |
|---|---|---|---|
| Cardiac surgery | |||
| IDU | 13.0 (7.5–21.5) | 9.8 (6.2–15.0) | −33% |
| Non-IDU | 16.3 (10.4– 24.7) | 12.4 (8.0–18.8) | −31% |
| Mortality | |||
| IDU | 3.7 (2.2–6.2) | 3.2 (1.8–5.4) | −16% |
| Non-IDU | 4.6 (2.7–7.8) | 4.0 (2.3–6.8) | −16% |
Notably, women (AOR 0.65, 95% CI 0.6–0.7), individuals from the lowest zip code income quartile (AOR 0.89, 95% CI 0.84–0.96), and those not admitted electively (AOR 0.7, 95% CI 0.66–0.73) were less likely to receive surgery throughout the study period (Table 2). Individuals with hospitalizations paid with commercial insurance were more likely to receive surgery (AOR 1.32, 95% CI 1.2–1.45).
Mortality
In the preintervention period, the in-hospital mortality rates were 7.9% (95% CI 6.8%–9.0%) for IDU-IE and 10.1% (95% CI 9.5%–10.7%) for non–IDU-IE hospitalizations. In the postintervention period, the proportions of hospitalizations resulting in death were 7.8% (95% CI 7.0%–8.6%) of IDU-IE and 9.7% (95% CI 9.2%–10.2%) of non–IDU-IE cases (Supplementary Figure 2). In adjusted segmented regression models following backwards selection (Supplementary Table 4), the odds of in-hospital mortality during the preintervention period decreased by 2% per quarter for both IDU-IE and non–IDU-IE cases (AOR 0.97, 95% CI 0.97–0.99; Table 2). Individuals with IDU-IE were 21% (AOR 0.79, 95% CI 0.71–0.89) less likely to experience in-hospital mortality throughout the study, compared to those with non–IDU-IE. Compared to projected preintervention trends, 2 years after the implementation of public reporting, the in-hospital mortality rate for IDU-IE cases changed from 3.7% (95% CI 2.2%–6.2%) to 3.2% (95% CI 1.8%–5.4%), and the rate for the non-IDU group changed from 4.6% (95% CI 2.7%–7.8%) to 4.0% (95% CI 2.3%–6.8%), a decrease of 16% for both groups (Figure 1; Table 3).
Sensitivity Analyses
Including observations with missing values (Supplementary Table 5) and excluding elective admissions (Supplementary Tables 6 and 7) did not substantively change the results. Additionally, surgical rates were similar when limiting the outcome to aortic valve surgeries, rather than all valve surgeries (Supplementary Table 6). When stratifying by surgery, the odds of mortality declined after public reporting for those who received valve surgery, but there was no change in the in-hospital mortality rate or difference by IDU status for those who did not receive surgery (Supplementary Table 7). Among those who received surgery, the mean mortality risk and illness severity scores decreased (3.43 to 3.24 [P < .001] and 3.67 to 3.64 [P < .001], respectively), while scores increased (2.86 to 3.37 [P < .001] and 3.36 to 3.41 [P < .001]) for those who did not receive surgery. When we varied the quarter for trend changes in segmented regression analyses, allowing for an assessment of the delayed effect of the intervention, the main models had the best or equivalent fit, compared with other quarters (Supplementary Table 8). Removing hepatitis C from the algorithm used to identify IDU-IE did not affect the results (Supplementary Tables 6 and 7). Finally, the percentage of hospitalizations resulting in transfers did not change over time, and including transfers in the cohort did not significantly change the results (Supplementary Figure 3; Supplementary Table 9).
DISCUSSION
The introduction of public reporting of aortic valve surgery outcomes in 2013 was associated with decreased odds of valve surgery for all patients with endocarditis. We did not detect a differential change by IDU status, suggesting that reporting did not worsen bias towards patients with IDU-IE [15, 16]. To our knowledge, this study is the first to demonstrate that public reporting measures were associated with a change in the odds of valve surgery for patients with endocarditis. Because individuals with endocarditis have higher mortality rates than those requiring valve surgery for other indications, reporting may have unintentionally resulting in fewer guideline indicated procedures, but higher mortality rates.
The main threat to the validity of ITS is co-occurring interventions; we are not aware of other changes in 2013 that would account for the observations. While fentanyl entered the heroin supply, causing overdose deaths in some regions in 2013, it is unlikely that it explains national surgical trends. A 2012 trial showed a benefit of early valve surgery in endocarditis, potentially mitigating within-group differences following reporting [12]. Additionally, Medicaid expansion in January 2014 potentially increased access to surgery. We observed increased surgery rates in hospitalizations paid by Medicaid after the intervention, and controlled for insurance type in our analyses. The NIS cannot be used for state estimates, so we were unable to account for state differences in the Medicaid expansion.
The in-hospital mortality rates were stable for both IDU-IE and non–IDU-IE patients in the baseline period. Following public reporting, there was a significant decrease in the odds of in-hospital mortality for endocarditis. The sensitivity analysis stratifying data by receipt of valve surgery showed decreased illness severity and mortality risk scores, as well as decreased odds of in-hospital mortality, among those who received surgery for IE following public reporting, compared to the baseline period. Among those who did not receive surgery, there was no change detected in in-hospital mortality or illness severity rates, but mortality risk scores increased. Further studies should investigate whether the increase in mortality risk scores was due to surgical case selection following reporting. Our examination of mortality was limited to the hospitalization, and the impact of decreased surgery on mortality may be experienced in a longer time frame: further studies should assess postdischarge mortality rates.
Our study’s strengths contribute to understanding surgical decision-making in endocarditis. We used nationally representative data over 5.75 years to identify and control for baseline trends for outcomes. In addition, we conducted the analysis at the individual level, using marginal effects to adjust for changes in patient characteristics, which is an important threat to validity in an ITS design. Our estimates of change between IDU and IDU-IE are strengthened given the comparator, protecting against nonobserved confounding [38] and modeling differences between groups in the baseline period. The results were similar after multiple sensitivity analyses to address missing data, misclassification by hepatitis C status, elective admission, and transfers, and to assess only aortic valve surgery. Additionally, we varied the intervention time and did not find delayed effects, perhaps due to an anticipation of policy implementation.
This observational study is subject to several limitations, in addition to the threat of co-occurring interventions. First, the attribution of patients to IDU-IE and non–IDU-IE is subject to misclassification, despite a validated algorithm to distinguish between IDU and non-IDU groups [33]. Coding for opioid use disorder accelerated following the introduction of ICD-10, so we excluded Q4/2015 in the analysis [39]. Second, we defined valve surgery using ICD-9 codes [40]. Some individuals with device infections may have been coded as endocarditis and received interventions absent from our valve surgery measure. Likely concentrated in the non-IDU group, this may underestimate procedures in non–IDU-IE cases, minimizing differences in between-group comparisons. Third, we were unable to obtain granular clinical information, including the organism or valve infected, or confirm that the valve surgery did not precede endocarditis. We excluded prosthetic valve endocarditis codes from the cohort, and postoperative prosthetic valve endocarditis in the immediate postoperative period is rare, with 0.99 cases in 100 person-years at 1 year [41]. Finally, we were unable to assess whether a patient had surgical indications, underwent surgery in a subsequent admission, or died after the hospitalization.
Our results have significant implications for policy-makers considering public reporting as a means to improve quality and clinicians caring for individuals with endocarditis. While public reporting may have improved the overall aortic valve surgery quality [25], the measures may have decreased access to valve surgery for IDU-IE and non–IDU-IE patients. IDU-IE infections result in significant morbidity and mortality for a relatively young population, and valve surgery may prevent subsequent mortality and disability, especially if clinicians screen for and deliver medications for opioid use disorder, like methadone and buprenorphine, to treat underlying substance use disorders [7, 17]. This policy was also associated with decreased valve surgeries for non–IDU-IE patients, who have more comorbidities. Although the AVR public reporting algorithm controlled for endocarditis, this may have been insufficient to impact valve surgery rates for this high-risk population. Public reporting may have negative consequences, which could be mitigated with a robust risk adjustment or the exclusion of those at the highest risk of complications [21].
CONCLUSION
In a nationally representative sample of inpatients, the implementation of public reporting of aortic valve surgery outcomes was associated with significant decreases in valve surgery rates for patients with IDU-IE and non–IDU-IE. This finding may represent an unintended consequence of public reporting.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Acknowledgments. The authors thank Ellen McCarthy and Shimon Shaykevich for assistance with data manipulation and Davida Schiff for thoughtful comments on this manuscript.
Potential conflicts of interest. S. D. K. was supported by the American Society of Addiction Medicine 2017 Annual Fellowship Award; the National Institute on Drug Abuse (NIDA), including the Clinical Addiction Research and Education Unit and Fellows Immersion Training Program (grant number R25DA013582) and the Research in Addiction Medicine Scholars Program (grant number R25DA033211); and the National Institute of Allergy and Infectious Diseases, through the Boston University Clinical Human Immunodeficiency Virus (HIV)/Acquired Immunodeficiency Syndrome (AIDS) Training Program (grant number 5T32AI052074). A. Y. W. received support from the Clinical Addiction Research and Education Unit (grant number R25DA013582). B. P. L. received support from the NIDA, including Researching Effective Strategies to Prevent Opioid Death (grant number R01DA046527); the Center on Health Economics of Substance Use Disorders, Hepatitis C Virus, and HIV Treatment in the Era of Integrated Health Care (grant number P30 DA040500); and the Providence/Boston Center for AIDS Research (grant number P30 AI042853). L. W. received support from the Providence/Boston Center for AIDS Research (grant number P30 AI042853). M. L. was supported by the NIDA (grant number K23 DA042168) and a Boston University School of Medicine Department of Medicine Career Investment Award; reports research support paid to his institution from OptumLabs; and received a grant from the Robert Wood Johnson Foundation, outside the submitted work. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
Reference
- 1. Ronan MV, Herzig SJ. Hospitalizations related to opioid abuse/dependence and associated serious infections increased sharply, 2002-12. Health Aff 2016; 35:832–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Wurcel AG, Anderson JE, Chui KKH, et al. Increasing infectious endocarditis admissions among young people who inject drugs. Open Forum Infect Dis 2016; 3:ofw157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Fleischauer AT, Ruhl L, Rhea S, Barnes E. Hospitalizations for endocarditis and associated health care costs among persons with diagnosed drug dependence — North Carolina, 2010–2015. MMWR Morb Mortal Wkly Rep 2017; 66:569–573. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Schranz AJ, Fleischauer A, Chu VH, Wu L-T, Rosen DL. Trends in drug use–associated infective endocarditis and heart valve surgery, 2007 to 2017: a study of statewide discharge data. Ann Intern Med 2019; 170:31–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Bor DH, Woolhandler S, Nardin R, Brusch J, Himmelstein DU. Infective endocarditis in the U.S., 1998–2009: a nationwide study. PLOS One 2013; 8:e60033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Baddour LM, Wilson WR, Bayer AS, et al. ; American Heart Association Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease of the Council on Cardiovascular Disease in the Young, Council on Clinical Cardiology, Council on Cardiovascular Surgery and Anesthesia, and Stroke Council Infective endocarditis in adults: diagnosis, antimicrobial therapy, and management of complications: a scientific statement for healthcare professionals from the American Heart Association. Circulation 2015; 132:1435–86. [DOI] [PubMed] [Google Scholar]
- 7. Rodger L, Glockler-Lauf SD, Shojaei E, et al. Clinical characteristics and factors associated with mortality in first-episode infective endocarditis among persons who inject drugs. JAMA Netw Open 2018; 1:e185220. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Tleyjeh IM, Abdel-Latif A, Rahbi H, et al. A systematic review of population-based studies of infective endocarditis. Chest 2007; 132:1025–35. [DOI] [PubMed] [Google Scholar]
- 9. Kiefer T, Park L, Tribouilloy C, et al. Association between valvular surgery and mortality among patients with infective endocarditis complicated by heart failure. JAMA 2011; 306:2239–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Lalani T, Cabell CH, Benjamin DK, et al. ; International Collaboration on Endocarditis-Prospective Cohort Study (ICE-PCS) Investigators Analysis of the impact of early surgery on in-hospital mortality of native valve endocarditis: use of propensity score and instrumental variable methods to adjust for treatment-selection bias. Circulation 2010; 121:1005–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Duval X, Delahaye F, Alla F, et al. Temporal trends in infective endocarditis in the context of prophylaxis guideline modifications: Three successive population-based surveys. J Am Coll Cardiol 2012; 59:1968–76. [DOI] [PubMed] [Google Scholar]
- 12. Kang DH, Kim YJ, Kim SH, et al. Early surgery versus conventional treatment for infective endocarditis. N Engl J Med 2012; 366:2466–73. [DOI] [PubMed] [Google Scholar]
- 13. Vlahakes GJ. “Consensus guidelines for the surgical treatment of infective endocarditis”: the surgeon must lead the team. J Thorac Cardiovasc Surg 2017; 153:1259–60. [DOI] [PubMed] [Google Scholar]
- 14. Hull SC, Jadbabaie F. When is enough enough? The dilemma of valve replacement in a recidivist intravenous drug user. Ann Thorac Surg 2014; 97:1486–7. [DOI] [PubMed] [Google Scholar]
- 15. Kim JB, Ejiofor JI, Yammine M, et al. Surgical outcomes of infective endocarditis among intravenous drug users. J Thorac Cardiovasc Surg 2016; 152:832–841.e1. [DOI] [PubMed] [Google Scholar]
- 16. Hall R, Shaughnessy M, Boll G, et al. Drug-use and post-operative mortality following valve surgery for infective endocarditis: a systematic review and meta-analysis. Clin Infect Dis 2018. Available at: https://academic.oup.com/cid/advance-article/doi/10.1093/cid/ciy1064/5257358 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Rosenthal ES, Karchmer AW, Theisen-Toupal J, Castillo RA, Rowley CF. Suboptimal addiction interventions for patients hospitalized with injection drug use-associated infective endocarditis. Am J Med 2016; 129:481–5. [DOI] [PubMed] [Google Scholar]
- 18. Springer SA, Korthuis PT, Del Rio C. Integrating treatment at the intersection of opioid use disorder and infectious disease epidemics in medical settings: a call for action after a National Academies of Sciences, Engineering, and Medicine workshop. Ann Intern Med 2018; 169:335–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Trowbridge P, Weinstein ZM, Kerensky T, et al. Addiction consultation services - linking hospitalized patients to outpatient addiction treatment. J Subst Abuse Treat 2017; 79:1–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Lemaire A, Dombrovskiy V, Saadat S, et al. Patients with infectious endocarditis and drug dependence have worse clinical outcomes after valvular surgery. Surg Infect 2017; 18:299–302. [DOI] [PubMed] [Google Scholar]
- 21. Shahian DM, He X, Jacobs JP, et al. The society of thoracic surgeons isolated aortic valve replacement (AVR) composite score: a report of the STS quality measurement task force. Ann Thorac Surg 2012; 94:2166–71. [DOI] [PubMed] [Google Scholar]
- 22. Svensson LG, Adams DH, Bonow RO, et al. Aortic valve and ascending aorta guidelines for management and quality measures. Ann Thorac Surg 2013; 95:S1–66. [DOI] [PubMed] [Google Scholar]
- 23. Badhwar V, Rankin JS, He X, et al. The society of thoracic surgeons mitral repair/replacement composite score: a report of the society of thoracic surgeons quality measurement task force. Ann Thorac Surg 2016; 101:2265–71. [DOI] [PubMed] [Google Scholar]
- 24. Brown JM, O’Brien SM, Wu C, Sikora JA, Griffith BP, Gammie JS. Isolated aortic valve replacement in North America comprising 108,687 patients in 10 years: changes in risks, valve types, and outcomes in the society of thoracic surgeons national database. J Thorac Cardiovasc Surg 2009; 137:82–90. [DOI] [PubMed] [Google Scholar]
- 25. D’Agostino RS, Jacobs JP, Badhwar V, et al. The society of thoracic surgeons adult cardiac surgery database: 2017 update on outcomes and quality. Ann Thorac Surg 2017; 103:18–24. [DOI] [PubMed] [Google Scholar]
- 26. O’Brien SM, Shahian DM, Filardo G, et al. The society of thoracic surgeons 2008 cardiac surgery risk models: part 2—isolated valve surgery. Ann Thorac Surg 2009; 88:S23–S42. [DOI] [PubMed] [Google Scholar]
- 27. The Society of Thoracic Surgery. About the STS Risk Calculators, Data Version 2.81. 2017. Available at: http://riskcalc.sts.org/stswebriskcalc/views/About the STS Risk Calculator v2 81.pdf Accessed 29 October 2018. [Google Scholar]
- 28. Campanella P, Vukovic V, Parente P, Sulejmani A, Ricciardi W, Specchia ML. The impact of public reporting on clinical outcomes: a systematic review and meta-analysis. BMC Health Serv Res 2016; 16:296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Blumenthal DM, Valsdottir LR, Zhao Y, et al. A survey of interventional cardiologists’ attitudes and beliefs about public reporting of percutaneous coronary intervention. JAMA Cardiol 2018; 3:629–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Gupta A, Yeh RW, Tamis-Holland JE, et al. Implications of public reporting of risk-adjusted mortality following percutaneous coronary intervention: misperceptions and potential consequences for high-risk patients including nonsurgical patients. JACC Cardiovasc Interv 2016; 9:2077–85. [DOI] [PubMed] [Google Scholar]
- 31. Romano PS, Marcin JP, Dai JJ, et al. Impact of public reporting of coronary artery bypass graft surgery performance data on market share, mortality, and patient selection. Med Care 2011; 49:1118–25. [DOI] [PubMed] [Google Scholar]
- 32. Omoigui NA, Miller DP, Brown KJ, et al. Outmigration for coronary bypass surgery in an era of public dissemination of clinical outcomes. Circulation 1996; 93:27–33. [DOI] [PubMed] [Google Scholar]
- 33. Ball LJ, Sherazi A, Laczko D, et al. Validation of an algorithm to identify infective endocarditis in people who inject drugs. Med Care 2018; 56:e70–e75. [DOI] [PubMed] [Google Scholar]
- 34. Roth J. Centers for Medicare and Medicard services general equivalence mapping 2012. Available at: http://www.nber.org/data/icd9-icd-10-cm-and-pcs-crosswalk-general-equivalence-mapping.html. Accessed 18 March 2018.
- 35. Averill RF, McCullough EC, Goldfield N, Hughes JS, Bonazelli J, Bentley L, Muldoon J, Turner L.. 3M APR DRG classification system: methodology overview Available at: https://www.hcup-us.ahrq.gov/db/nation/nis/grp031_aprdrg_meth_ovrview.pdf. Accessed 22 February 2019.
- 36. Leary RS, Johantgen ME, Farley D, Forthman MT, Wooster LD. All-payer severity-adjusted diagnosis-related groups: a uniform method to severity-adjust discharge data. Top Health Inf Manage 1997; 17:60–71. [PubMed] [Google Scholar]
- 37. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther 2002; 27:299–309. [DOI] [PubMed] [Google Scholar]
- 38. Bernal JL, Cummins S, Gasparrini A. Interrupted time series regression for the evaluation of public health interventions: A tutorial. Int J Epidemiol 2017; 46:348–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Heslin KC, Owens PL, Karaca Z, Barrett ML, Moore BJ, Elixhauser A. Trends in opioid-related inpatient stays shifted after the us transitioned to ICD-10-CM diagnosis coding in 2015. Med Care 2017; 55:918–23. [DOI] [PubMed] [Google Scholar]
- 40. Peterson C, Xu L, Mikosz CA, Florence C, Mack KA. US hospital discharges documenting patient opioid use disorder without opioid overdose or treatment services, 2011–2015. J Subst Abuse Treat 2018; 92:35–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Glaser N, Jackson V, Holzmann MJ, Franco-Cereceda A, Sartipy U. Prosthetic valve endocarditis after surgical aortic valve replacement. Circulation 2017; 136:329–31. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.