Although studies have demonstrated health benefits, there is limited evidence on utilization and cost changes associated with cardiac pacemaker implantation from national community samples. The aim of this study was to quantify changes in emergency room (ER) and hospital inpatient use and in Medicare payments per beneficiary/year after pacemaker implantation. Outcomes for pacemaker recipients after and before implantation and between pacemaker recipients and controls were compared using propensity score matching. Data came from Health and Retirement Study interviews merged with Medicare claims. Sample subjects were aged ≥68 years with diagnosed conduction disorders or cardiac dysrhythmias in the previous 3 years. Outcome measures were (1) ER visits, inpatient admissions and days, and Medicare payments for ER and inpatient care in the after period for the pacemaker versus control groups, defined per beneficiary/year, (2) difference in differences in the same 5 outcome variables, and (3) binary variables for whether or not utilization or payments were lower in the after versus before periods for the pacemaker versus control groups. In conclusion, most pacemaker recipients improved, as measured by reductions in use and payments in the after versus before period, and there were reductions in ER visits and hospital admissions for conditions commonly leading to pacemaker implantation.
Cardiac pacemakers have existed for decades, but their use is increasing and will increase further because of population aging and other factors. Most pacemaker recipients are ≥65 years of age. Although studies have demonstrated health benefits, there is very limited empirical evidence on utilization and cost changes associated with pacemaker receipt. There are studies of pacemaker outcomes using data from randomized controlled trials, but longitudinal evidence from national community samples is lacking. In this study, we used a national sample of elderly patients merged with Medicare claims to assess changes in emergency room (ER) and hospital inpatient use and in payments per beneficiary/year for such services after pacemaker receipt.
Methods
Data came from the Health and Retirement Study (HRS), a longitudinal survey of subjects aged 51 to 61 years in 1992 and their spouses or partners of any age, with older and younger cohorts added subsequently. The HRS has been conducted in even-numbered years since 1992. The HRS elicits information on personal health, physical and cognitive function, income, and other topics. Merged with HRS data, Medicare claims provided information on diagnoses, utilization, and Medicare payments for services beneficiaries received. We used HRS interview data to match pacemaker recipients with controls. Our sample consisted of patients diagnosed with conduction disorders (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 426) and cardiac dysrhythmias (ICD-9-CM code 427), recorded as primary or secondary diagnoses during a 3-year look-back period. Given the look-back period length, all sample subjects were ≥68 years of age.
Pacemaker implantation procedures were for the implantation of cardiac resynchronization pacemaker devices without mention of defibrillation, total system (Current Procedural Terminology, 4th Edition, code 0050), or implantation of pacemaker systems (single, dual, biventricular with leads) in patients without preexisting pacemakers or leads (Healthcare Common Procedure Coding System codes 33206, 33207, and 33208). The observational period was from 1996 to 2004, during which 434 sample patients received pacemakers. The HRS provided data on health and functional status, demographic characteristics, and earnings and income. HRS data were obtained periodically, while claims were reported daily. Thus, we split dates of service on Medicare claims into discrete time periods to match claims with HRS interview years. For the pacemaker sample, a period consisted of claims with pacemaker implantation dates 6 months before to 6 months after the HRS interview year. For example, we matched pacemaker implantations occurring from July 1, 1997, to June 30, 1999, with 1998 HRS data and implantations from July 1, 1999, to June 30, 2001, with 2000 HRS data.
We observed utilization and Medicare payments, the study outcomes, before and after pacemaker implantation, for beneficiaries receiving pacemakers. Having a control group allowed us to account for changes independent of pacemaker implantation, such as regression toward the mean in utilization or payments. If subjects experience unusually high or low utilization in a year, there may be convergence toward the mean value for the group subsequently. Control groups consisted of beneficiaries with diagnoses of conduction disorders or cardiac dysrhythmias. Initially, we selected all beneficiaries with these diagnoses from 1996 to 2004 claims (n = 13,602). Controls entered the study on the basis of the first claim with the study diagnoses from 1996 to 2004. The date of this claim was the “reference date” from which look-back and look-forward data were calculated.
We compared changes in the utilization of ER and inpatient care and Medicare payments between a period 1 year before pacemaker implantation and, for controls, a 1-year period immediately before the reference date and for a year after a 6-month period after pacemaker placement or the reference date (controls). The 6-month gap from implantation to follow-up allowed adjustments to be made after implantation (e.g., to device settings and medications). Outcome measures defined per beneficiary/year were ER visits, inpatient admissions and days, and Medicare payments for ER and inpatient care.
Although the analysis sample used a common set of diagnoses, pacemaker recipients may differ on characteristics unobservable to researchers. To reduce selection bias in assigning subjects to the intervention, we used propensity score matching (PSM) to obtain a control group to compare utilization and Medicare payments after pacemaker placement.
To implement PSM, we first performed logit analysis to predict the probability a beneficiary received a pacemaker. Using the predicted probability, we matched a beneficiary actually receiving a pacemaker to his or her nearest match among controls. We then measured utilization and payment outcome differences for pacemaker recipients versus controls. We used nearest neighbor matching with a caliper of 0.02, with PSMATCH2 from Stata version 11 (StataCorp LP, College Station, Texas). Observation pairs were dropped if differences in values exceeded this amount. Standardized differences were calculated for the matched sample. A general criterion for adequate matching is that standardized differences for the covariates used for matching not exceed 10%.
We performed logit analysis and PSM using these covariates from HRS: demographic characteristics: age (75 to 84 or ≥85 years, with 68 to 74 years omitted), race or ethnicity (black or Hispanic, white or other omitted), female gender, currently married, and years of schooling (12, 13 to 15, or ≥16 years, with <12 years omitted); health and functional status: binary variables for subjects with fair or poor self-reported health and obesity (body mass index ≥30 kg/m 2 ), number of limitations in activities of daily living (difficulty or inability to walk across a room, dress by oneself, use the toilet, bathe, and get out of bed), number of limitations in instrumental activities of daily living (difficulty or inability to shop and prepare meals), and mobility (difficulty or inability to walk several blocks, walk 1 block, get out of a chair, climb 1 flight of stairs, climb several flights of stairs, stoop, carry an object weighing 10 lb, pick up a dime, and push large object); annual earnings ($5,000 to <$50,000 or ≥$50,000, with <$5,000 omitted); and annual household income other than from the respondent’s earnings ($5,000 to <$50,000, $50,000 to <$100,000, or ≥$100,000, with <$5,000 omitted).
From Medicare claims, we included binary variables for heart failure (HF; ICD-9-CM codes 428.0x to 428.4x, 428.9x, 398.91, 402.02, 402.11, and 402.91), coronary artery disease (ICD-9-CM codes 410.xx to 414.xx and no HF diagnosis), and diabetes mellitus (ICD-9-CM codes 250.xx) during the past 3 years. We also included a binary variable for beneficiaries whose most recent “before” claim with an ICD-9-CM code 426.xx or 427.xx diagnosis, before pacemaker implantation for the pacemaker and before the reference date for the control group, was for sinoatrial node dysfunction, the code for bradycardia (ICD-9-CM code 427.8) and a covariate for pacemaker implantation year and for controls, the reference year, and utilization or Medicare payments during the previous year. The utilization or payment explanatory variable depended on the outcome being analyzed. We used a 1-year look-back for utilization and Medicare payments, inflated to 2011 dollars using the Consumer Price Index, to match the 1-year period during follow-up.
We present average treatment effects on treated subjects (ATTs) for matched samples. The ATT measures the effects of a treatment among those receiving the treatment (the pacemaker group) versus no treatment (the control group); PSMATCH2 gives the ATT as the difference in mean values between treatment and control groups at follow-up and associated standard errors, from which t values and statistical significance are calculated.
We computed ATTs in 3 ways. We measured the difference in use and payments in the after period between the pacemaker and control groups. The ATT represented the difference in mean values of dependent variables at follow-up between pacemaker and controls. Alternatively, for a difference-in-differences estimate, we first computed the difference in outcomes in the follow-up versus before periods for the pacemaker and control groups separately. Then we subtracted the difference in outcomes (follow-up minus before) for controls from the difference in outcomes for the pacemaker group. Third, we measured whether utilization or payments for the pacemaker group decreased (improved). Improvements were set to 1 and lack of improvements to 0. The ATT was the difference in 2 binary variables: improvement for the pacemaker group minus improvement for controls. Thus, the ATT was the fraction of beneficiaries improving among pacemaker recipients relative to the fraction improving for controls.
Results
Before matching, the largest differences in values between the pacemaker and control groups as indicated by standardized differences were for diagnosis of bradycardia in the before period ( Table 1 ): 41% in the pacemaker group versus 2% in the control group. Pacemaker recipients were more likely to be men and older than controls and had much higher annual payments for ER and hospital inpatient care in the previous year; 28% of pacemaker recipients were diagnosed with HF versus 18% of controls. Pacemaker recipients had better functional status. The groups were similar in socioeconomic factors. When such comparisons were possible, characteristics of the pacemaker sample were similar to those reported in other pacemaker studies.
| Variable | Before Matching | After Matching | ||||
|---|---|---|---|---|---|---|
| Pacemaker | Control | Standardized Difference | Pacemaker | Control | Standardized Difference | |
| Black | 0.10 | 0.12 | −3.95 | 0.11 | 0.13 | −6.50 |
| Hispanic | 0.07 | 0.05 | 7.25 | 0.06 | 0.07 | −3.15 |
| Female | 0.45 | 0.56 | −20.9 | 0.47 | 0.40 | 14.3 |
| Married | 0.54 | 0.51 | 4.55 | 0.53 | 0.57 | −7.91 |
| Age 75–84 yrs | 0.50 | 0.42 | 16.5 | 0.50 | 0.49 | 0.52 |
| Age ≥85 yrs | 0.22 | 0.20 | 6.37 | 0.22 | 0.23 | −2.52 |
| Education 12 yrs | 0.27 | 0.32 | −11.5 | 0.28 | 0.27 | 1.77 |
| Education 13–15 yrs | 0.15 | 0.16 | −3.55 | 0.16 | 0.15 | 1.45 |
| Education ≥16 yrs | 0.17 | 0.15 | 3.49 | 0.17 | 0.18 | −4.16 |
| Fair/poor health | 0.46 | 0.44 | 3.38 | 0.45 | 0.49 | −7.36 |
| Body mass index | 0.19 | 0.17 | 4.14 | 0.18 | 0.22 | −9.24 |
| Number of limitations of activities of daily living (0–5) | 0.26 | 0.32 | −12.7 | 0.27 | 0.26 | 2.38 |
| Number of limitations of instrumental activities of daily (0–2) | 0.20 | 0.24 | −10.6 | 0.21 | 0.18 | 5.95 |
| Number of mobility limitations (0–9) | 0.84 | 0.85 | −4.01 | 0.83 | 0.84 | −4.23 |
| Heart failure | 0.28 | 0.18 | 24.3 | 0.25 | 0.30 | −10.6 |
| Coronary artery disease | 0.03 | 0.06 | −11.4 | 0.04 | 0.04 | −1.37 |
| Diabetes mellitus | 0.15 | 0.13 | 8.35 | 0.15 | 0.14 | 2.25 |
| Bradycardia | 0.41 | 0.02 | 108.1 | 0.35 | 0.32 | 7.21 |
| Earned $5,000 to <$50,000 | 0.06 | 0.05 | 2.37 | 0.05 | 0.08 | −10.6 |
| Earned ≥$50,000 | 0.01 | 0.01 | −5.16 | 0.00 | 0.00 | 0.00 |
| Income $5,000 to <$50,000 | 0.82 | 0.81 | 1.56 | 0.81 | 0.79 | 4.56 |
| Income $50,000 to <$100,000 | 0.12 | 0.12 | −1.16 | 0.12 | 0.15 | −6.90 |
| Income ≥$100,000 | 0.05 | 0.05 | 3.53 | 0.05 | 0.05 | 1.19 |
| ER visits/beneficiary/year | 1.80 | 0.80 | 50.9 | 1.70 | 1.87 | −5.59 |
| ER payments/beneficiary/year (×1,000) | 9.98 | 2.22 | 69.4 | 9.19 | 9.36 | −0.89 |
| Number of inpatient admissions | 1.43 | 0.49 | 76.6 | 1.35 | 1.35 | −0.30 |
| Number of inpatient days | 10.3 | 3.39 | 60.3 | 9.74 | 9.37 | 2.17 |
| Inpatient payments/1,000 | 18.0 | 4.44 | 74.9 | 17.2 | 18.0 | −2.70 |
| Year | 6.68 | 5.83 | 29.6 | 6.57 | 6.89 | −11.1 |
| n | 434 | 13,602 | 381 | 381 | ||
Such patterns were also observed in multivariate logit analysis ( Table 2 ). Only PSM results for the inpatient admissions analysis are listed. Results for other outcome variables were very similar. Women and subjects with more limitations in activities of daily living and instrumental activities of daily living were much less likely to receive pacemakers. Age, previous HF diagnosis, and a larger number of inpatient admissions in the before year were positively associated with pacemaker receipt. A claim with sinoatrial node dysfunction diagnosis (bradycardia) immediately before pacemaker receipt made pacemaker receipt much more likely. The odds ratio for year (1.12) implies an important trend in pacemaker implantation.
| Variable | Odds Ratio | 95% Confidence Interval |
|---|---|---|
| Black | 0.95 | 0.65–1.37 |
| Hispanic | 1.09 | 0.68–1.75 |
| Female | 0.72 ∗ | 0.56–0.91 |
| Married | 0.99 | 0.77–1.28 |
| Age 75–84 yrs | 1.97 ∗ | 1.50–2.58 |
| Age ≥85 yrs | 2.43 ∗ | 1.71–3.44 |
| Education 12 yrs | 0.76 | 0.57–1.01 |
| Education 13–15 yrs | 0.84 | 0.59–1.19 |
| Education ≥16 yrs | 0.90 | 0.62–1.30 |
| Fair/poor health | 1.06 | 0.83–1.36 |
| Body mass index | 1.33 | 0.99–1.78 |
| Number of limitations of activities of daily living (0–5) | 0.65 ∗ | 0.48–0.89 |
| Number of limitations of instrumental activities of daily (0–2) | 0.59 ∗ | 0.42–0.84 |
| Number of mobility limitations (0–9) | 0.75 | 0.55–1.04 |
| Heart failure | 1.35 ∗ | 1.04–1.77 |
| Coronary artery disease | 0.63 | 0.35–1.15 |
| Diabetes mellitus | 1.14 | 0.83–1.56 |
| Bradycardia | 38.7 ∗ | 30.0–49.89 |
| Earned $5,000 to <$50,000 | 1.48 | 0.91–2.40 |
| Earned ≥$50,000 | 0.20 | 0.03–1.66 |
| Income $5,000 to <$50,000 | 1.37 | 0.60–3.12 |
| Income $50,000 to <$100,000 | 1.33 | 0.54–3.27 |
| Income ≥$100,000 | 1.65 | 0.62–4.42 |
| Number of inpatient admissions/beneficiary/year | 1.67 ∗ | 1.56–1.80 |
| Year | 1.12 ∗ | 1.08–1.17 |
After matching, pacemaker group observations decreased from 434 to 373 to 384, depending on the outcome ( Table 1 ). Observations were dropped from the original pacemaker sample of 434 because adequate matches were not found given our matching criteria. After matching, almost all standardized differences were <10%.
The pacemaker group experienced decreases in use and payments after versus before pacemaker implantation ( Table 3 ). Rates per beneficiary/year decreased by 0.59 for ER visits, by 0.64 for inpatient admissions, by 4.51 for inpatient days, by $5,547 for Medicare payments for ER services, and by $10,221 for payments for inpatient care. All ATTs for differences in differences were not statistically significant, because there were comparable decreases for controls. There were also no statistically significant ATT values for levels of outcomes. However, almost half or more of the pacemaker recipients improved on each outcome. Only about a third of controls did, and all ATTs were statistically significant. The largest ATT, 0.29, was for improvements (reductions) in inpatient days per year per beneficiary. The lowest relative improvement was 0.11 for ER visits per year per beneficiary.
| Variable | Pacemaker | Control | ATT | 95% Confidence Interval |
|---|---|---|---|---|
| Differences in outcomes after vs before pacemaker implantation | ||||
| ER visits (768 observations) | −0.59 | −0.62 | 0.04 | −0.35 to 0.42 |
| ER payments ($)/1,000 (764 observations) | −5.55 | −5.17 | −0.38 | −3.07 to 2.31 |
| Number of inpatient admissions (746 observations) | −0.64 | −0.63 | −0.01 | −0.31 to 0.28 |
| Number of inpatient days (762 observations) | −4.51 | −3.86 | −0.65 | −3.38 to 2.09 |
| Inpatient payments ($)/1,000 (762 observations) | −10.2 | −12.2 | 1.94 | −2.67 to 6.55 |
| Outcomes after pacemaker implantation | ||||
| ER visits (768 observations) | 1.12 | 1.25 | −0.13 | −0.49 to 0.24 |
| ER payments ($)/1,000 (764 observations) | 3.65 | 4.19 | −0.54 | −2.01 to 0.92 |
| Number of inpatient admissions (746 observations) | 0.71 | 0.72 | −0.02 | −0.23 to 0.19 |
| Number of inpatient days (762 observations) | 5.23 | 5.51 | −0.28 | −2.33 to 1.77 |
| Inpatient payments ($)/1,000 (762 observations) | 6.96 | 5.85 | 1.10 | −1.02 to 3.23 |
| Improvement in outcomes after vs before pacemaker implantation | ||||
| ER visits (768 observations) | 0.49 | 0.38 | 0.11 ∗ | 0.04 to 0.18 |
| ER payments ($)/1,000 (764 observations) | 0.58 | 0.36 | 0.22 ∗ | 0.15 to 0.29 |
| Number of inpatient admissions (746 observations) | 0.54 | 0.37 | 0.16 ∗ | 0.09 to 0.23 |
| Number of inpatient days (762 observations) | 0.6 | 0.32 | 0.29 ∗ | 0.22 to 0.36 |
| Inpatient payments ($)/1,000 (762 observations) | 0.62 | 0.39 | 0.24 ∗ | 0.17 to 0.31 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
