Association of Early Outpatient Rehabilitation With Health Service Utilization in Managing Medicare Beneficiaries With Nontraumatic Knee Pain: Retrospective Cohort Study

Abstract

Background. Nontraumatic knee pain (NTKP) is highly prevalent in adults 65 years of age and older. Evidence-based guidelines recommend early use of rehabilitation; however, there is limited information comparing differences in health care utilization when rehabilitation is included in the management of NTKP.

Objectives. To describe the overall health care utilization associated with the management of NTKP; estimate the proportion of people who receive outpatient rehabilitation services; and evaluate the timing of outpatient rehabilitation and its association with other health care utilization.

Design. Rretrospective cohort study was conducted using a random 10% sample of 2009–2010 Medicare claims. The sample included 52,504 beneficiaries presenting within the ambulatory setting for management of NTKP.

Methods. Exposure to outpatient rehabilitative services following the NTKP index ambulatory visit was defined as 1) no rehabilitation; 2) early rehabilitation (1-15 days); 3) intermediate rehabilitation (16-120 days); and 4) late rehabilitation (>120 days). Logistic regression models were fit to analyze the association of rehabilitation timing with narcotic analgesic use, utilization of nonsurgical invasive procedure, and knee surgery during a 12-month follow-up period.

Results. Only 11.1% of beneficiaries were exposed to outpatient rehabilitation services. The likelihood of using narcotics, nonsurgical invasive procedures, or surgery was significantly less (adjusted odds ratios; 0.67, 0.50, 0.58, respectively) for those who received early rehabilitation when compared to no rehabilitation. The exposure-outcome relationships were reversed in the intermediate and late rehabilitation cohorts.

Limitations. This was an observational study, and residual confounding could affect the observed relationships. Therefore, definitive conclusions regarding the causal effect of rehabilitation exposure and reduced utilization of more aggressive interventions cannot be determined at this time.

Conclusions. Early referral for outpatient rehabilitation may reduce the utilization of health services that carry greater risks or costs in those with NTKP.

Knee pain and symptomatic knee osteoarthritis, which can be generally categorized as nontraumatic knee pain (NTKP), are prevalent and associated with greater disability in people 65 years of age and older.¹ It has been reported that among the elderly these conditions have increased 65% over a 20 year period, with rates doubling in women and tripling in men.² Not surprising, this trend coincides with an increase in the rate of total knee replacement surgeries, adding significantly to the cost of managing people with knee pain.

Several evidence-based guidelines for the treatment of NTKP recommend rehabilitation (exercise therapy or other nonpharmacological interventions) as an early stage intervention to reduce pain and disability, and to potentially limit the need for total knee arthroplasty (TKA).^3-6 This is an especially important topic given current policy efforts to address the opioid epidemic and curb the escalating cost of TKA.^7,8 Unfortunately, recent reports suggest that many people with NTKP are not receiving outpatient rehabilitation services as part of the management plan.^9,10

If receipt of rehabilitation is beneficial in reducing pain and disability, it would seem that people who receive these services might exhibit less utilization of more aggressive medical interventions such as narcotics, knee injections, arthroscopic surgery, and total knee arthroplasty. However, there is limited information comparing the health services utilization associated with managing NTKP between people who receive rehabilitation and those who do not. It is important to explore whether exposure to outpatient rehabilitation services reduces the utilization of potentially risky or costly procedures over time. This finding could impact guideline implementation and policy strategies designed to optimize access to—and use of—rehabilitation for people with NTKP.

The aims of this study were to: 1) describe the health care utilization associated with the management of NTKP in people 65 years and older; 2) determine the proportion of people with NTKP who receive outpatient rehabilitation services as part of their management and describe the timing of those services; and 3) examine the impact of rehabilitation on health care service utilization for managing knee pain.

Methods

Data Sources

The data for this retrospective study came from a 10% random sample of Medicare claims covering dates of service from January 1, 2009, to December 31, 2010. These data provided utilization and cost information for individual beneficiaries linked across the following research identifiable files:

• MedPar (ie, acute care hospitals and skilled nursing facilities);

• Outpatient (eg, rehabilitation agencies, ambulatory surgical centers);

• Home Health Agency;

• Carrier (ie, medical professionals); and

• Part D Drug Event

The dataset did not include claims from inpatient rehabilitation facilities.

NTKP Cohort Identification

The study cohort included Medicare beneficiaries 65 years and older with NTKP as defined by 19 International Classification of Diseases, 9th Revision (ICD-9) diagnosis codes¹¹ (Appendix  1). To identify individuals presenting for the evaluation or treatment of a new episode of NTKP, or those returning after an extended 6-month break in care, the 24 months of claims data was divided into 3 distinct periods: 1) a 6-month run-in period; 2) a 6-month cohort identification period; and 3) a 12-month run-out period. The data were divided in this fashion to allow sufficient time to aggregate an adequate number of beneficiaries with an episode of interest, and also to ensure a uniform 12-month follow-up period for all beneficiaries in the sample. (Appendix  2) In an effort to mitigate the possibility of misclassification and loss to follow-up biases, all beneficiaries had to be continuously enrolled in Medicare Part A and B throughout 2009–2010.

To be considered for inclusion in the analytic cohort, a beneficiary had to have at least one date of service with one of the 19 NTKP diagnoses in the primary position on an ambulatory claim (eg, professional office or outpatient visit) during the 6-month cohort identification period. The earliest event meeting these criteria was considered the “index event.” Additionally, the beneficiary's index event had to be immediately preceded by a 6-month “clean period,” in which there were no claims with a primary NTKP diagnosis identified in any hospital, skilled nursing, home health, or ambulatory care setting during the run-in period. When this last condition was met, all knee-related events regardless of setting were aggregated according to the beneficiary's unique identification number for the entire 12-month period (ie, run-out) immediately following the index event (Appendix  2). The final NTKP cohort contained 52,504 beneficiaries meeting these criteria (Figure).

Figure.

Cohort construction flow diagram. NTKP = nontraumatic knee pain.

Health Care Utilization Categories

To succinctly describe the utilization associated with the management of NTKP, we categorized both the medical and pharmacy claims into clinically similar categories (eAppendix, available at academic.oup.com/ptj). Medical utilization was identified using Health Care Common Procedure Coding System (HCPCS) Level 1 (CPT-4) and 2, Revenue Center, and ICD-9 Procedure codes. Both professional and facility claims were included. In some instances, a single claim line had 2 types of utilization codes (ie, Revenue Center and CPT-4) available in the raw dataset. Where this occurred, a hierarchical algorithm was used to ensure that we categorized the service using the code that provided the most detailed description. Our categorization of medical utilization resulted in 9 categories: rehabilitation, knee surgery, nonsurgical invasive procedures (eg, injections, aspiration), medical (eg, physician office visits), pharmacy, radiology, laboratory, other medical, and other therapy.¹¹

Rehabilitation was operationally defined as exercise and other nonpharmacological services or procedures which are recommended as early stage management options for patients with NTKP.^3-6 Included in this category were various forms of exercise, nutritional counseling, functional training, physical agents, manipulation, and manual therapy. Inclusion in this category was dependent on neither the type of provider rendering the service, nor the setting in which they were performed.

Additionally, pharmacy utilization was further categorized into 9 subcategories: anticonvulsants, antidepressants, anxiolytics, glucocorticoids, histamine-2 blockers, nonsteroidal anti-inflammatories (NSAID), narcotic analgesics, proton pump inhibitors, and sedatives.¹¹ It should be noted that prescription drug data carried in the Medicare Part D file does not contain ICD-9 diagnosis codes. Therefore, we employed an attribution method to identify the use of a prescription drug associated with the treatment of NTKP. First, the prescription had to fall within a drug class (eg, NSAIDs) commonly used in the treatment of NTKP. To determine which classes of medications to include in our analysis, we used the medication list from a large randomized trial in a similar population in which patients self-reported the medications they were taking for NTKP (Clinical Trials.gov NCT01314183). Second, the prescription drug service date, or the date that the drug was dispensed, had to be temporally associated with the NTKP episode and fall within the 12-month follow-up period. All prescription drug claims meeting these criteria were included in the analysis.

Dependent Variables

We analyzed 3 dichotomous indicators (ie, Yes/No) of health care exposure associated with the management of NTKP. The 3 exposure variables were: indicators of any use of narcotic analgesics, nonsurgical invasive procedures, and knee surgery during the 12-month follow-up period.

Independent Variable–Timing of Rehabilitation Exposure

The primary independent variable was a 4-level categorical variable calculated based on the number of days between the index event and the date of the first rehabilitative service(s) as: 1) no rehabilitation; 2) early rehabilitation; 3) intermediate rehabilitation; and 4) late rehabilitation. A beneficiary was considered to have early exposure if any rehabilitation was initiated within 15 days of the index event; intermediate exposure was any rehabilitation starting between 16–120 days from the index event; and exposure was considered late if rehabilitation was initiated more than 120 days from the index event. For a non-trivial portion of rehabilitation users (33%), their only exposure to rehabilitative services was following surgery. Because the clinical indications, precautions, progressions, and goals for rehabilitation are considerably different after knee surgery, beneficiaries were not considered “exposed” to rehabilitation if their only use of these services was postsurgical.

Control Variables

Three categories of control variables were used: demographic characteristics, geographic location, and clinical comorbidities. The demographic characteristics included sex, age, and race. Age was treated as a continuous variable but was rescaled so that 1 unit change equaled 5 years. Race was categorized as white, black, Hispanic, Asian, North American Native, other, and unknown. Geographic location consisted of the standard 4 US Census regions (Northeast, Midwest, South, and West).

The clinical comorbidities were based on the Functional Comorbidity Index (FCI). The FCI was originally developed as a self-report instrument designed to capture the presence/absence of 18 health conditions associated with changes in physical function.¹² The conditions used in the index are: arthritis, osteoporosis, asthma, chronic obstructive pulmonary disease, angina, congestive heart failure, myocardial infarction, neurologic disease, stroke, peripheral vascular disease, diabetes, upper gastrointestinal disease, depression, anxiety, visual impairment, hearing impairment, degenerative disc disease, and obesity. We used England et al's crosswalk between the FCI comorbidities and their corresponding ICD-9 diagnostic codes to identify these conditions in the claims files.¹³ A dichotomous indicator was created for each of the 18 conditions. A beneficiary was considered to have the comorbidity if there was an event identified with an applicable diagnosis in any position on any claim within the period for which we had claims (ie, 24 months).

Data Analysis

Stata Release 13.1 (College Station, TX) was used for all analyses. We calculated descriptive statistics for beneficiary demographic characteristics, geographic location, and clinical comorbidities. Summary statistics of health care utilization were computed by medical and pharmacy utilization categories, as well as the timing of rehabilitation exposure. These statistics included the count and proportion of users. For the 3 dependent variables, we fit multivariable logistic regression models to evaluate the association of rehabilitation timing on our measures of health care utilization. These models were fit using a forced, backward stepwise approach. The full model was fit to include the rehabilitation timing variable and all control variables. We included each clinical comorbidity in the models individually as opposed to using the summated FCI, because this approach explained more variance.¹⁴ Control variables were manually removed from the model 1 at a time if their P value exceeded .10. This was done until the most parsimonious model was identified as confirmed by the likelihood ratio test. The odds ratios from the multivariable logistic models reflect the net association between the timing of rehabilitation exposure and the dependent variables, after holding the final set of control variables constant in the model. We considered odds ratios with P values ≤ .05 to be statistically significant.

Results

The analytic file included 52,504 beneficiaries who met our definition of NTKP and were followed for 12 months. The distribution of beneficiaries according to each demographic characteristic, geographic location, and clinical comorbidity variable is presented in Table 1. The majority of beneficiaries were female (72.6%) and white (79.2%). The mean age was 76.6 years. The South region had the highest proportion of beneficiaries with NTKP (38.8%), followed by the Midwest (24.8%), Northeast (19.5%), and West (16.9%). Five of the 18 clinical comorbidities were present in at least one-third of the sample. These were arthritis (82.4%), congestive heart failure (54.9%), visual impairment (48.1%), degenerative disc disease (43.6%), and diabetes (36.3%).

Table 1.

Descriptive Statistics for Medicare Beneficiaries 65 Years of Age and Older With Nontraumatic Knee Pain^a

			95% CI
	Sample Size	Percentage or Mean	Lower	Upper
Overall	52,504	100.0%
Gender
Female	38,120	72.6%	72.2%	73.0%
Male	14,384	27.4%	27.0%	27.8%
Age (years)	52,504	76.6	76.6	76.7
Race
White	41,563	79.2%	78.8%	79.5%
Black	4,386	8.4%	8.1%	8.6%
Hispanic	4,063	7.7%	7.5%	8.0%
Asian	1,821	3.5%	3.3%	3.6%
Other	396	0.8%	0.7%	0.8%
North American Native	233	0.4%	0.4%	0.5%
Unknown	42	0.1%	0.1%	0.1%
Census Region
South	20,367	38.8%	38.4%	39.2%
Midwest	13,029	24.8%	24.4%	25.2%
Northeast	10,237	19.5%	19.2%	19.8%
West	8,870	16.9%	16.6%	17.2%
Comorbidities
Arthritis	43,238	82.4%	82.0%	82.7%
CHF or heart disease	28,847	54.9%	54.5%	55.4%
Visual impairment	25,257	48.1%	47.7%	48.5%
Degenerative disc disease	22,906	43.6%	43.2%	44.1%
Diabetes	19,045	36.3%	35.9%	36.7%
Upper GI disease	15,910	30.3%	29.9%	30.7%
Osteoporosis	10,619	20.2%	19.9%	20.6%
Depression	8,315	15.8%	15.5%	16.2%
Stroke/TIA	8,004	15.2%	14.9%	15.6%
Peripheral vascular disease	6,393	12.2%	11.9%	12.5%
Anxiety/panic disorder	6,251	11.9%	11.6%	12.2%
Asthma	5,167	9.8%	9.6%	10.1%
COPD, ARDS, emphysema	5,033	9.6%	9.3%	9.8%
Hearing impairment	3,952	7.5%	7.3%	7.8%
Myocardial infarction	3,859	7.3%	7.1%	7.6%
Obesity	3,745	7.1%	6.9%	7.4%
Angina	2,795	5.3%	5.1%	5.5%
Neurological disease	1,319	2.5%	2.4%	2.6%

^aCI = confidence interval, CHF = congestive heart failure, TIA = transient ischemic attack, COPD = chronic obstructive pulmonary disease, ARDS = acute respiratory distress syndrome.

Health Care Utilization

Of the 9 health service utilization categories, the estimated proportion of use was highest in the Pharmacy (84%), Medical (75.8%), Radiology (67.3%), and Nonsurgical Invasive procedure (40.2%) categories. In our sample, 8.7% of beneficiaries had knee surgery and 16.5% had outpatient rehabilitation services (Table 2).

Table 2.

Health Care Resources Utilization for Medicare Beneficiaries 65 Years of Age and Older With Nontraumatic Knee Pain by Medical Category^a

		95% CI			95% CI
Medical Category	User Count	Lower	Upper	Percentage	Lower	Upper
Pharmacy	44,121	43,956	44,286	84.0%	83.7%	84.3%
Medical	39,799	39,607	39,991	75.8%	75.4%	76.2%
Radiology	35,345	35,134	35,556	67.3%	66.9%	67.7%
Other Medical	30,921	30,700	31,142	58.9%	58.5%	59.3%
Nonsurgical Invasive	21,130	20,910	21,350	40.2%	39.8%	40.7%
Rehabilitation	8,672	8,505	8,839	16.5%	16.2%	16.8%
Surgical	4,566	4,439	4,693	8.7%	8.5%	8.9%
Laboratory	2,159	2,070	2,248	4.1%	3.9%	4.3%
Other Therapy	224	195	253	0.4%	0.4%	0.5%

^aCI = confidence interval.

Timing of Rehabilitation Exposure

There were 8672 beneficiaries who received outpatient rehabilitation services for NTKP during the 12-month follow-up period. As stated early, the services for a large percentage of these users (32.5%) were strictly limited to postsurgical care. Because the clinical indications and goals for rehabilitation are considerably different after knee surgery, these beneficiaries were not considered “exposed” to rehabilitation. This resulted in a total of 5852 beneficiaries, or 11.1% of the total sample, who were exposed to outpatient rehabilitation services as part of NTKP management. In the exposure group, 52% received outpatient rehabilitation services within the first 15 days of the index event (early); 27% had outpatient rehabilitation services 16–120 days from the index event (intermediate); and 21% had late exposure where outpatient rehabilitation services were initiated more than 120 days from the index event (Table 3).

Table 3.

Rehabilitation Status of Medicare Beneficiaries 65 Years of Age and Older with Nontraumatic Knee Pain

	Full Sample			All Rehabilitation Usersa			Rehabilitation Exposure of Interestb
	(n = 52,504)			(n = 8,672)			(n = 5,852)
		95% CId			95% CId			95% CId
	Percentage	Lower	Upper	Percentage	Lower	Upper	Percentage	Lower	Upper
No Rehabilitation	83.5%	88.6%	89.1%
Postsurgical Rehabilitation Only	5.4%	5.2%	5.5%	32.5%	31.5%	33.5%
Early Rehabilitation Exposures	5.8%	5.6%	6.0%	35.2%	34.2%	36.3%	52.2%	50.9%	53.5%
Intermediate Rehabilitation Exposurec	3.0%	2.8%	3.1%	18.1%	17.3%	18.9%	26.8%	25.7%	27.9%
Late Rehabilitation Exposurec	2.3%	2.2%	2.5%	14.2%	13.5%	14.9%	21.0%	20.0%	22.1%
	100.0%			100.0%			100.0%

^aIncludes all beneficiaries who were rehabilitation users regardless of clinical indication.

^bRehabilitation users exposed to services of interest.

^cTiming of rehabilitation exposure from index visit: Early (0–15 days), Intermediate (16–120 days), Late (>120 days).

^dCI = confidence interval.

Narcotic Analgesic Utilization

The adjusted odds for the use of narcotic analgesics in beneficiaries with NTKP when exposed to early rehabilitation is 0.67 (95%CI, 0.62-0.72) times lower when compared to those who did not receive outpatient rehabilitation services. This same relationship does not hold true for beneficiaries who receive intermediate or late rehabilitation exposure. In the intermediate and late exposure groups, the adjusted odds of narcotic use are higher when compared to the reference group; OR 1.43 (95%CI, 1.28-1.60) and OR 1.32 (95%CI, 1.16-1.49), respectively. There was also significant racial variation in narcotic use. When compared to Whites, the adjusted odds of narcotic use were higher for Blacks (OR 1.31, 95%CI, 1.22-1.40) and North American Natives (OR 1.80, 95%CI, 1.34-2.42). However narcotic use was lower for Asians (adjusted OR 0.48, 95%CI, 0.44-0.54) and Hispanics (adjusted OR 0.79, 95%CI, 0.74-0.85). The adjusted odds of narcotic use were significantly higher in the Midwest (OR 1.55, 95%CI, 1.47-1.64), South (OR 1.84, 95%CI, 1.75-1.93), and West (OR 1.75, 95%CI, 1.65-1.86) when compared to the Northeast. Each of the 5 most prevalent comorbidities were significantly associated with narcotic utilization: Arthritis (adjusted OR 1.47, 95%CI, 1.40-1.54), CHF/Heart Disease (adjusted OR 1.31, 95%CI, 1.26-1.36), Visual Impairment (adjusted OR 0.85, 95%CI, 0.82-0.88), Degenerative Disc Disease (adjusted OR 1.96, 95%CI, 1.89-2.04), and Diabetes (adjusted OR 1.20, 95%CI, 1.15-1.25) (Table 4).

Table 4.

Results from Multiple Logistic Regression Models for Timing of Rehabilitation Exposure and Utilization of Narcotic Medications, Nonsurgical Invasive Procedures, and Surgical Procedures in Medicare Beneficiaries 65 Years of Age and Older with Nontraumatic Knee Pain^a

	Narcotic Use Model			Nonsurgical Invasive Use Model			Surgical Use Model
		95% CI			95% CI			95% CI
	Odds Ratio	Lower	Upper	Odds Ratio	Lower	Upper	Odds Ratio	Lower	Upper
Rehabilitation Timing
No Rehabb	–	–	–	–	–	–	–	–	–
Early Rehab (<16 days)	0.67	0.62	0.72	0.50	0.46	0.55	0.58	0.49	0.69
Intermediate Rehab (16–120 days)	1.43	1.28	1.60	1.47	1.32	1.63	1.39	1.19	1.62
Late Rehab (>120)	1.32	1.16	1.49	2.12	1.88	2.39	1.03	0.85	1.25
Gender
Maleb	–	–	–				–	–	–
Female	1.16	1.12	1.21	Variable omittedc			0.82	0.77	0.88
Age (one unit equals 5 years)	0.92	0.90	0.93	0.98	0.97	1.00	0.76	0.74	0.78
Race
Whiteb	–	–	–	–	–	–	–	–	–
Black	1.31	1.22	1.40	0.61	0.57	0.65	0.48	0.41	0.55
Other	0.68	0.55	0.84	0.92	0.74	1.14	0.68	0.46	1.00
Asian	0.48	0.44	0.54	0.70	0.63	0.78	0.35	0.27	0.46
Hispanic	0.79	0.74	0.85	0.61	0.56	0.65	0.54	0.47	0.62
North American Native	1.80	1.34	2.42	0.59	0.44	0.80	0.64	0.37	1.09
Unknown	0.59	0.31	1.12	0.60	0.30	1.18	0.88	0.26	2.97
Census Region
Northeastb	–	–	–	–	–	–	–	–	–
Midwest	1.55	1.47	1.64	1.13	1.06	1.19	1.61	1.45	1.78
South	1.84	1.75	1.93	1.29	1.22	1.36	1.41	1.28	1.56
West	1.75	1.65	1.86	0.93	0.87	0.99	1.65	1.47	1.84
Comorbidities
Arthritis	1.47	1.40	1.54	7.13	6.66	7.64	27.23	20.00	37.09
CHF or heart disease	1.31	1.26	1.36	Variable omittedc			1.38	1.29	1.48
Visual impairment	0.85	0.82	0.88	1.11	1.07	1.15	Variable omittedc
Degenerative disc disease	1.96	1.89	2.04	1.07	1.03	1.11	0.80	0.75	0.85
Diabetes	1.20	1.15	1.25	0.93	0.89	0.97	0.84	0.78	0.90
Upper GI disease	1.32	1.27	1.38	1.04	1.00	1.08	1.31	1.23	1.40
Osteoporosis	Variable omittedc			0.93	0.89	0.97	Variable omittedc
Depression	1.40	1.33	1.48	0.83	0.79	0.88	0.90	0.82	0.99
Stroke/TIA	Variable omittedc			Variable omittedc			0.85	0.77	0.94
Peripheral vascular disease	1.10	1.03	1.16	0.83	0.78	0.88	0.83	0.74	0.93
Anxiety/panic disorder	1.24	1.16	1.32	Variable omittedc			Variable omittedc
Asthma	1.18	1.11	1.26	Variable omittedc			1.32	1.19	1.46
COPD, ARDS, emphysema	1.24	1.16	1.33	0.84	0.79	0.90	0.70	0.62	0.80
Hearing impairment	0.87	0.81	0.93	Variable omittedc			0.84	0.73	0.96
Myocardial infarction	1.19	1.11	1.29	Variable omittedc			Variable omittedc
Obesity	1.43	1.33	1.55	Variable omittedc			1.88	1.71	2.08
Angina	Variable omittedc			1.08	0.99	1.17	0.78	0.67	0.91
Neurological disease	Variable omittedc			0.73	0.64	0.82	0.53	0.41	0.70

^aCI = confidence interval, CHF = congestive heart failure, TIA = transient ischemic attack, COPD = chronic obstructive pulmonary disease, ARDS = acute respiratory distress syndrome.

^bReference Group.

^cForced backward elimination used to eliminate variable when P > .10.

Nonsurgical Invasive Procedure Utilization

The adjusted odds of NTKP beneficiaries using nonsurgical invasive procedures when exposed to early rehabilitation was 0.50 (95%CI, 0.46-0.55) times lower when compared to those who do not receive outpatient rehabilitation services. This relationship is reversed in beneficiaries receiving intermediate or late rehabilitation exposure. The adjusted odds of nonsurgical invasive procedure utilization in the intermediate (OR 1.47, 95%CI, 1.32-1.63) and late exposure (OR 2.12, 95%CI, 1.88-2.39) groups are significantly higher. The adjusted odds of receiving nonsurgical invasive procedures is significantly less for all race groups when compared to Whites, including Blacks (OR 0.61, 95%CI, 0.57-0.65), North American Natives (OR 0.59, 95%CI, 0.44-0.80), Asians (OR 0.70, 95%CI, 0.63-0.78), and Hispanics (OR 0.61, 95%CI, 0.56-0.65), respectively. The adjusted odds of nonsurgical invasive procedure use is highest in the South region (OR 1.29, 95%CI, 1.22-1.36) and lowest in the West region (OR 0.93, 95%CI, 0.87-0.99). The adjusted odds of receiving a nonsurgical invasive procedure are 7.13 (95%CI, 6.66-7.64) times higher in beneficiaries diagnosed with arthritis when compared to those without the condition (Table 4).

Knee Surgery Utilization

The adjusted odds for beneficiaries with NTKP having knee surgery when exposed to early rehabilitation is 0.58 (95%CI, 0.49-0.69) times lower when compared to those who do not receive outpatient rehabilitation services. The adjusted odds of females having knee surgery are 0.82 (95%CI, 0.77-0.88) times lower than males. The adjusted odds of having knee surgery is significantly less for Blacks (OR 0.48, 95%CI, 0.41-0.55), Asians (OR 0.35, 95%CI, 0.27-0.46), and Hispanics (OR 0.54, 95%CI, 0.47-0.62), when compared to Whites. The adjusted odds of having knee surgery were significantly higher in the Midwest (OR 1.61, 95%CI, 1.45-1.78), South (OR 1.41, 95%CI, 1.28-1.56), and West (OR 1.65, 95%CI, 1.47-1.84) when compared to the Northeast. The adjusted odds of having surgery were 1.88 (95%CI, 1.71-2.08) times higher in beneficiaries diagnosed with obesity and 27.23 (95%CI, 20.00-37.09) times higher for those with arthritis when compared to beneficiaries without these conditions (Table 4).

Discussion

The study cohort consisted of 52,504 Medicare beneficiaries who received evaluation or management services for a primary complaint of NTKP. In our sample, approximately 17% of beneficiaries received outpatient rehabilitation services; however, for one-third of these beneficiaries these services were provided solely for postsurgical rehabilitation and were not used as treatment for NTKP per se. This means that the percentage of beneficiaries who received any rehabilitation was approximately 11%. This figure is consistent with previously published findings in similar patient populations.¹⁵ This is important, considering our finding that early use of outpatient rehabilitation services (ie, within 15 days of the index visit) in the NTKP population is associated with lower use of narcotic analgesics, injections, and surgery when compared to beneficiaries not exposed to rehabilitation. Therefore, developing strategies to encourage the use of rehabilitation as a first-line treatment for NTKP, as recommended by current guidelines, has the potential to positively impact a large segment of this clinical population.

The pharmacologic management of acute and chronic pain in older adults is complex. Nonsteroidal anti-inflammatory (NSAIDs) medications, a common therapeutic approach for musculoskeletal disorders, carry additional risk in older adults and evidence-based guidelines recommend that they be used judiciously.^16,17 As a result, many practitioners choose to control pain in the older adult using narcotic analgesics. However, these pharmacologic interventions also increase the risk of adverse events such as dependency, falls, fractures, and mortality.^16,18,19 In our sample, 57% of beneficiaries with NTKP filled one or more prescriptions for a narcotic analgesic. However, the likelihood of filling a narcotic prescription during the follow-up period was significantly lower for patients exposed to early rehabilitation as compared to those that did not receive rehabilitation (adjusted OR = 0.67) or those who received it later in the course of care. Our findings would seem to support the recent recommendations that nonpharmacological treatment options, including those delivered by physical therapists, should be considered prior to treatment with narcotic prescription.⁷

Clinical guidelines generally recommend that nonsurgical invasive procedures (eg, corticosteroid injections) are to be used only after simple pharmacologic and non-pharmacologic interventions are no longer efficacious in managing pain and disability associated with NTKP.¹⁵ Adverse events associated with these procedures can range from minor skin reactions to more serious complications such as sepsis.²⁰ While the risk of serious complications is small, these unintended consequences should be an important consideration given their potential seriousness.¹⁵ It would appear that the practice patterns in our cohort are contrary to this recommendation, as we saw a high rate of use (40%) of nonsurgical invasive services and a much lower rate of rehabilitation use (11%) in patients with NTKP. Furthermore, patients received nonsurgical invasive services much sooner following the index visit (mean 39.4 days 95%CI, 33.9-35.9) compared to outpatient rehabilitation services (mean 63.6 days 95%CI, 61.2-66.0). This sequencing appears to be important because for patients exposed to early rehabilitation, the likelihood of receiving nonsurgical invasive services was cut in half (adjusted OR 0.50) when compared to beneficiaries who did not receive rehabilitation exposure. Conversely, patients who received intermediate or late rehabilitation exposure had a much higher likelihood of being treated with nonsurgical invasive procedures (adjusted ORs 1.47 and 2.12, respectively). Once more, it appears that early rehabilitation may be a valuable management option for NTKP patients who are poor candidates for, or prefer to forego, nonsurgical invasive treatment procedures.

Finally, the sharp rise in surgical procedures, especially total knee arthroplasty, in NTKP patients has been well documented.²¹ In response Medicare has implemented a Comprehensive Care for Joint Replacement bundled payment model in an effort to improve quality and control the cost of TKA.⁸ However, this effort does not address the frequency of TKAs. Some investigators analyzing the rising TKA rates have opined that there is overuse of these services and that up to one-third of these procedures are inconsistent with clinical guideline recommendations.^22,23 Reasons cited include a poor match between the condition and the surgical procedure, and a failure to exhaust conservative care measures before proceeding to a surgical intervention.^22,23 Two findings from our study are particularly pertinent given these observations. First, only 11% beneficiaries who underwent surgery had any outpatient rehabilitation services in the 6 months immediately preceding their procedure. Second, when exposed to early rehabilitation, beneficiaries were much less likely [adjusted OR = 0.58] to progress to surgery over the 12-month follow-up period compared to those beneficiaries who were not exposed. Given the apparent benefit of early rehabilitation, at least in the short term, it would appear reasonable to exhaust this treatment option before moving on to these more invasive, riskier, and expensive interventions.

The results of our study are intriguing, but we acknowledge that they should be considered preliminary at this time. It is important to keep in mind the limitations inherent in observational research studies. Because patients were not randomly assigned to rehabilitation treatments, we cannot draw conclusions about the causal effect of rehabilitation exposure and reduced utilization of more aggressive interventions. As a result, there may be unmeasured confounders that could strengthen or diminish the observed relationships in all, or some, of the 3 rehabilitation groups. For example, one could conclude from our findings that no rehabilitation is superior to intermediate or late rehabilitation. However, it is important to consider that this relationship may be affected by differences in patient characteristics across these groups. Claims data do not contain clinical markers of pain or function. It is reasonable to suspect that intermediate or late rehabilitative services would tend to be used in patients exhibiting higher levels of pain and disability, or those with persistent symptoms, which could potentially explain their higher use of narcotic analgesics, injections, and surgery. Likewise, the protective effect of early rehabilitation could be positively or negatively affected by pain and function characteristics of the individuals, or by other unmeasured confounding present at the time services are delivered. This reality points to the importance of ongoing efforts to link clinical measures with health care service utilization from claims data to identify subgroups of patients who may benefit best from early vs delayed rehabilitation and/or from rehabilitation services in general.

In this study, it would also have been ideal to precisely isolate beneficiaries with incident NTKP to allow initial management decisions to be more precisely studied. Our dataset only contained 2 years of health care claims. This gave us a limited run-in period (6 months) to identify previous treatment for NTKP; therefore, we were unable to determine with certainty where beneficiaries were in the long-term course of their care for this progressive condition. Another limitation of claims data is the variation in the billing codes used by different provider specialties and in different treatment settings. This variation in coding precluded a detailed assessment of the content and dosage of specific outpatient rehabilitation services. Nevertheless, even without detail about specific treatments, our findings suggest that mere exposure to a broadly defined set of rehabilitation strategies is associated with reductions in the use of potentially risky or costly procedures.

Finally, it is important to consider both resource utilization and cost. However, in the current study we were unable to generate accurate estimates of the total cost of care. In the NTKP population, surgery is the major driver of health care costs. Additionally, previous investigators have found that a large proportion of older adults receive postsurgical rehabilitation in inpatient rehabilitation facilities, which adds considerably to the total cost of care.^24,25 Unfortunately, our dataset did not include claims from inpatient rehabilitation facilities. These missing data did not affect analyses of resource utilization outcomes; however, we were unable to accurately estimate the total health care costs associated with changes in rehabilitation exposure. A cost of care analysis in this population will be a focus of our future work.

While we cannot make definitive conclusions from our data, our results suggest that finding ways to promote prompt access to outpatient rehabilitation services for people with knee pain may be important for reducing utilization of health services that place patients at greater risk. This specifically relates to opioid use, nonsurgical invasive procedures, and surgery. Rehabilitation appears to have its best effects when provided early in the course of care. Further study is needed to identify the best candidates for, and to test strategies to improve, early utilization of outpatient rehabilitation services, and to test the effects of these strategies on downstream health care utilization and costs.

Author Contributions and Acknowledgments

Concept/idea/research design: J.M. Stevans, G.K. Fitzgerald, S.R. Piva, M. Schneider

Writing: J.M. Stevans, G.K. Fitzgerald, S.R. Piva, M. Schneider

Data collection: J.M. Stevans

Data analysis: J.M. Stevans, G.K. Fitzgerald, S.R. Piva, M. Schneider

Project management: J.M. Stevans

Dr Fitzgerald is a Catherine Worthingham Fellow of the American Physical Therapy Association. Dr Piva is a Board-Certified Clinical Specialist in Sports Physical Therapy and a Fellow of the American Academy of Orthopaedic Manual Physical Therapists.

The authors acknowledge Todd Jahangiri of ePrecise Solutions Inc for his expert programming and data management skills, which were instrumental in the completion of this project.

Ethics Approval

This study was reviewed by the University of Pittsburgh Institutional Review Board and considered exempt (PRO12090068).

Disclosures and Presentations

The authors declare no potential conflicts of interest. Dr Stevans is a paid consultant to Landmark Healthplan of California. Dr Schneider provides medico-legal consulting services and expert testimony on chiropractic fraud and malpractice and is a member of the NCMIC Inc Speakers Bureau.

Appendix 1. ICD-9 Diagnostic Codes Used to Identify the Nontraumatic Knee Pain Cohort

Code	Description
715.16	Osteoarthrosis, localized, primary, lower leg
715.26	Osteoarthrosis, localized, secondary, lower leg
715.36	Osteoarthrosis, localized, not specified whether primary or secondary, lower leg
715.96	Osteoarthrosis, unspecified whether generalized or localized, lower leg
716.46	Transient arthropathy, lower leg
716.56	Unspecified polyarthropathy or polyarthritis, lower leg
716.66	Unspecified monoarthritis, lower leg
716.96	Arthropathy, unspecified, lower leg
717.7	Chondromalacia of patella
718.46	Contracture of joint, lower leg
719.46	Pain in joint
719.56	Stiffness of joint, not elsewhere classified
719.96	Unspecified disorder of joint
726.60	Enthesopathy of knee, unspecified
726.61	Pes anserinus tendinitis or bursitis
726.64	Patellar tendinitis
726.65	Prepatellar bursitis
726.69	Enthesopathy of knee, other
727.83	Plica syndrome

Appendix 2. Construction of the Nontraumatic Knee Pain Cohort