Neurologist-associated reduction in PD-related hospitalizations and health care expenditures

Standard protocol approvals, registrations, and patient consents.

This study was approved by the Human Research Protection Office at Washington University School of Medicine.

Study population.

The study population was derived from 100% Medicare part A and B outpatient and carrier claims data from the year 2002. All beneficiaries with a claim containing the International Classification of Diseases–9 code “332.0” were extracted for further analysis. Incident cases were identified using previously published methods.⁷ Beneficiaries who had concurrent diagnostic claims for “Secondary Parkinsonism” (332.1) or “Other Degenerative Diseases of the Basal Ganglia” (333.0), beneficiaries who were nursing home residents at the time of initial diagnosis (as indicated in the “place of service” variable),^8,9 and those who died prior to January 1, 2006, were excluded.

Individual demographic, comorbidity, and residential data were extracted from Medicare Beneficiary Annual Summary Files (BASFs). Demographic data included subject race, sex, and date of birth. Baseline clinical data on comorbid illness, defined as the presence of diabetes, dementia, malignant solid tumor (lung, colon/rectum, breast, prostate, uterine), ischemic heart disease, chronic obstructive pulmonary disease, cerebrovascular disease, acute myocardial infarction, chronic kidney disease, or congestive heart failure, were used to calculate an age-weighted Charlson comorbidity score.¹⁰ County of residence was linked to a previously developed socioeconomic deprivation score, comprised of census-based socioeconomic data: % high school graduate, % population unemployed, % population living in a crowded house, % population without a car, % population without a telephone, and % population below the federal poverty rate, weighted to reflect the community-level factors that may impact access to specialty care.¹¹

PD physician specialty.

We used provider identification codes to determine the treating physician specialty for each medical claim. Three neurologist involvement categories—“infrequent” (no outpatient neurologist PD visits), “intermediate” (1–5 neurologist outpatient PD visits), and “frequent” (more than 5 outpatient neurologist PD visits)—were created based on tertiles of the natural distribution of care over the 48-month observation period. The total number of outpatient PD visits, regardless of provider specialty, was calculated and included as a covariate in all analyses.

Hospitalization patterns according to neurologist frequency of care.

We hypothesized that persons with PD who received regular neurologist care would be less likely to be hospitalized for PD-related conditions compared to those who saw a neurologist infrequently for PD care. Hospitalization data were extracted for all patients from 2002 until December 31, 2005, using BASFs, which also contain diagnosis-related group (DRG) codes for each of the first 10 hospitalizations per year for each beneficiary. DRG codes are used to identify the principal diagnosis for a hospitalization for the purpose of standardizing billing and payments nationwide. We created DRG-based categories for hospitalization, choosing to examine 5 commonly reported disease-mediated or related medical conditions (referred to as “PD-related”) that may require hospitalization: psychosis, neurodegenerative disease, urinary tract infection, traumatic injury, and depression. These conditions were chosen a priori based on a review of the literature and our team's extensive clinical experience managing PD. The inclusion of hospitalization for depressive neuroses, the most commonly reported psychiatric illness in PD,^12–16 is straightforward. The other PD-related outcomes chosen deserve further explanation. Dysautonomia commonly produces bladder and bowel symptoms in PD (often prior to the classic tremor, slowness, shuffling, and poverty of movement), predisposing those with PD to urinary tract infection. Moreover, urinary tract infection has been shown to have an atypical presentation in PD: symptoms such as acute change in motor benefit from medications, confusion, and worsening of motor symptoms are common, as opposed to the urinary urgency, urinary frequency, and dysuria which are reported in typical urinary tract infections.^3–5,17 Psychosis in PD may be due to the disease itself, and occurs with increasing frequency as the disease progresses.^18–23 In a newly diagnosed patient with PD, however, psychosis is more likely secondary to PD medication regimen or polypharmacy. Therefore, we hypothesized that in this incident cohort, hospitalization patterns for psychosis may provide information on both adverse drug events risk and allow us to indirectly assess the baseline distribution of disease severity by physician specialty. Traumatic injury in the elderly, similar to hip fracture, is most likely due to falls. Fall prevention in PD relies on optimal motor management as well as appropriate use of physiatrist and rehabilitation services. We included the DRG code for “neurodegenerative disease” to determine if DRG coding for PD-related hospitalizations differed by physician specialty; i.e., if neurologist-treated patients were assigned a hospitalization code for “neurodegenerative disease” by default. For comparison, we extracted hospitalization data for 5 common general medical conditions that would trigger a hospitalization not likely to be prevented by neurologist care: angina pectoralis, diabetes mellitus, hypertension, gastrointestinal obstruction, and congestive heart failure.

Because initial hospitalization for a complication may be in part due to baseline patient health or disease severity rather than physician knowledge, training, or skill, we further hypothesized that neurologist care would also be associated with a lower likelihood of repeat hospitalization for PD-related illnesses, and compared the risk of additional hospitalization by neurologist treatment category, adjusting for the same confounders.

Finally, BASFs also contain cost data grouped by service type (outpatient, physician, inpatient, SNF, durable medical equipment, and home health) and yearly utilization data (number of hospitalizations, physician office visits, and covered SNF days). Total and subcategory cost and utilization data for 2002–2005 were extracted from these files for this cohort and compared by neurologist treatment category.

Statistical analyses.

Descriptive analyses performed on the cohort included the following: race, sex, age, physician specialty, and urban/rural residence classification. Cox proportional hazard models were used to explore the association of neurologist care frequency with the likelihood of PD-related and general medical illness hospitalizations. The time-to-event variable for each outcome was from the beginning of the study in 2002 until the first event occurrence (measured in years). Patients with no events were censored at the end of calendar year in 2005. Covariates included race, sex, age, socioeconomic deprivation, modified Charlson comorbidity index (CCI) score, and total number of outpatient PD visits. Primary models compared the “frequent” neurologist care category to the reference “infrequent” category. Additional analyses to evaluate for a dose-dependent reduction in PD-related hospitalization included the intermediate category of neurologic PD care in the Cox model. Finally, the risk of repeat hospitalization for each PD-related outcome was compared across neurologist treatment categories using logistic regression, adjusting for all previously indicated covariates. Means of continuous variables were compared by t test or analysis of variance. Proportions were evaluated using χ² tests. Logistic regression coefficients were converted to odds ratios with 95% confidence intervals (CIs) using standard methods. All statistical tests were 2-sided, with the statistical significance level set at 0.05, and with 95% CIs where appropriate. Statistical analyses were performed using SAS v.9.2, PASW v.16 and 17.

Subject demographics and baseline characteristics.

We identified 24,929 independently living beneficiaries with an incident PD diagnosis in 2002 who were still living on January 1, 2006. A greater proportion of Medicare recipients with PD who received frequent neurologist services were white or male and were slightly younger than Medicare recipients with PD who had infrequent neurologist care. The mean age-weighted modified Charlson comorbidity score differed slightly between those with infrequent and frequent neurologist PD care, 3.36 ± 0.70 vs 3.13 ± 0.67. The absolute differences in individual socioeconomic and geographic variables between those with frequent, intermediate, and infrequent neurologist care were also small (table 1).

Hospitalization patterns.

During the 48-month observation period, 5,318 (21.3%) patients with PD had at least 1 hospitalization for a PD-related illness. Adjusting for patient race, age, sex, modified Charlson comorbidity score, and socioeconomic deprivation, those with regular neurologist care had lower adjusted odds of being hospitalized for some PD-related illnesses: psychosis (hazard ratio [HR] 0.71, 95% CI 0.59–0.86), traumatic injury (HR 0.56, 95% CI 0.40–0.78), and urinary tract infection (HR 0.74, 95% CI 0.63–0.87) (table 2). There was no compensatory increased risk of hospitalization for neurodegenerative disease, nor was there an increase in risk of hospitalization for depression. The association between neurologist care and reduced risk of hospitalization appeared to be specific for PD-related comorbidities, as there was no difference in likelihood of hospitalization for the following common medical conditions across neurologist frequency groups: diabetes (HR 0.76, 95% CI 0.53–1.08), hypertension (HR 1.02, 95% CI 0.70–1.5), gastrointestinal obstruction (HR 1.01, 95% CI 0.76–1.30), congestive heart failure (HR 1.05, 95% CI 0.86–1.28), or angina (HR 1.10, 95% CI 0.75–1.62). Sensitivity analyses which included only those who had seen a neurologist at least once (for diagnosis confirmation) produced very similar results for both PD-related and general medical illnesses, without change in direction or magnitude of the observed associations.

Repeat hospitalization.

There were 1,433 repeat hospitalizations for PD-related illnesses in our cohort. Regular neurologist care was associated with a reduction in the likelihood that a person with PD would undergo more than 1 hospitalization for psychosis (adjusted odds ratio [OR] 0.41, 95% CI 0.33–0.51). The reduction in risk for repeat hospitalization for urinary tract infection (adjusted OR 0.50, 95% CI 0.35–0.71) was similar (table 3). There was no difference in the risk of repeat hospitalization for traumatic injury. Similar to our primary data, no difference in repeat hospitalization risk was found across neurologist treatment groups for hospitalization due to neurodegenerative disease, depression, or for the general medical illnesses we examined.

Utilization and cost patterns.

Total cost for PD patient care in our cohort was $1.48 billion during the 4-year observation period ($59,530 per person). Medicare-covered SNF days and hospitalizations were inversely related to neurologist involvement in PD care: there were an average of 27.5 (±49.4) SNF days and 3.2 (±2.9) hospitalizations among those in the infrequent neurologist treatment category and 15.2 (±33.6) SNF days and 2.73 (±2.3) hospitalizations among those in the frequent neurologist treatment category (each, p < 0.01, t test). Similarly, per person expenditures for inpatient, durable medical equipment, SNF, home health, and hospice care services were lowest for those with the most frequent neurologist care (table 4).