Abstract
Background
Musculoskeletal conditions are common and cause high levels of disability and costs. Physical therapy is recommended for many musculoskeletal conditions. Past research suggests that referral rates appear to have increased over time, but the rate of accessing a physical therapist appears unchanged.
Objective
Our retrospective cohort study describes the rate of physical therapy use after referral for a variety of musculoskeletal diagnoses while comparing users and non-users of physical therapy services after referral.
Methods
The study sample included patients in the University of Utah Health system who received care from a medical provider for a musculoskeletal condition. We included a comprehensive set of variables available in the electronic data warehouse possibly associated with attending physical therapy. Our primary analysis compared differences in patient factors between physical therapy users and non-users using Poisson regression.
Results
15 877 (16%) patients had a referral to physical therapy, and 3812 (24%) of these patients accessed physical therapy after referral. Most of the factors included in the model were associated with physical therapy use except for sex and number of comorbidities. The receiver operating characteristic curve was 0.63 suggesting poor predictability of the model but it is likely related to the heterogeneity of the sample.
Conclusions
We found that obesity, ethnicity, public insurance and urgent care referrals were associated with poor adherence to physical therapy referral. However, the limited predictive power of our model suggests a need for a deeper examination into factors that influence patients access to a physical therapist.
Keywords: Access to care, health services, musculoskeletal/connective tissue disorder, orthopaedics, patient adherence, physical therapy/physiotherapy
Key Messages.
Patients use PT after referral for musculoskeletal problems at a rate of 24%.
Patient factors influencing choice and access likely drive PT use rates.
PT use increased with orthopaedic referral and less distance to the clinic.
Obesity, using public payers, and Hispanic ethnicity reduced PT use.
Introduction
Musculoskeletal conditions are common and can cause high levels of disability for patients and high costs for health systems (1). There is a growing burden caused by musculoskeletal conditions driving a need to understand treatments that can reduce disability and improve outcomes (2,3). Concerns about increasing costs and overuse of opioid medications for musculoskeletal conditions have focussed attention on alternative, non-pharmacologic treatment options (4–6).
In common musculoskeletal conditions interventions performed by physical therapists are effective at improving pain and function (5–10). Given the effectiveness of many physical therapy treatments, several clinical guidelines recommend physical therapy for musculoskeletal conditions (1,5,11). While clinical guidelines and research appear to have increased rates of referral, there has not been a subsequent increase in the proportion of patients receiving care from a physical therapist (12–15). This difference is surprising as the traditional model of care expects patients will follow health care provider recommendations, thus an increase in referrals would result in an expected increase in use of physical therapy services (7). However, a complex interplay of factors drive access (12,16). These factors must be understood to comprehend health care use to improve care effectiveness.
Andersen’s Behavioral Model of Health Service use provides a framework to explore the factors that influence patients’ use of physical therapy services (17). Beyond patients with pelvic floor and back pain diagnoses, there is a lack of research exploring factors associated with compliance to a physical therapy referral. Limited studies on select conditions (back pain and pelvic floor pain) have identified self-navigation, patient factors and attitudes towards physical therapy as factors influencing physical therapy use (7,12,16). Few of these studies have identified cohorts from a common starting point such as receiving a provider recommendation, or referral, for physical therapy and are unable to differentiate the influence patient factors have on the use of physical therapy services. This study used a referral to physical therapy as a common starting point to examine the rate at which patients access physical therapy care. We aimed to identify factors differentiating users and non-users of physical therapy services in an integrated US health system for a variety of musculoskeletal diagnoses.
Methods
We conducted a retrospective cohort study of patients aged 18 to 64 who received a referral to physical therapy for a musculoskeletal condition at the University of Utah Health (UUH) system between 1 October 2015 and 30 September 2018.
Data source
We identified patients from the University of Utah’s Enterprise Data Warehouse (EDW), which contains administrative and clinical data from UUH in Salt Lake City, UT. The EDW includes clinical and demographic data of UUH patients, including electronic referrals, insurance and visit detail (provider type), International Classification of Diseases (ICD) 10 diagnosis codes, and Current Procedural Terminology codes (12).
Sample
The initial cohort included patients, ages 18–64, who had a visit with a UUH medical care provider for a new musculoskeletal diagnosis (referred to as the index visit). We included patients whose index visit occurred with physicians, nurse practitioners or physician assistants from pain management, urgent and emergency care, physiatry, primary care and orthopaedic departments/specialties. Providers from these departments often act as an entry point provider for patients suffering from ambulatory musculoskeletal diagnoses.
The musculoskeletal diagnosis codes included identified patients who are commonly seen by physical therapists (Supplemental 1) (12). Newly consulting patients were defined as having no previous musculoskeletal diagnosis for the same body region 6 months before their index visit, to ensure the patients are seeking care for a new problem including acute or exacerbated musculoskeletal disorders.
We excluded patients with any diagnoses suggesting possible non-musculoskeletal cause for their symptoms and those who were non-ambulatory or severely ill. Patients were excluded if they had any ICD-10 diagnosis codes representing cancer, infection, systemic disease or trauma in their record 4 weeks before or after their index visit. Non-ambulatory and severely ill patients were excluded if they had any diagnosis or referral codes indicating hemiplegia, paraplegia, quadriplegia, wheelchair dependence and end-stage renal disease 3 months before or after their index visit. We excluded individuals with a diagnosis of cauda equina within 6 months of their index visit. Patients who had a referral to physical therapy from infectious disease or cancer providers or had a referral diagnosis code indicating trauma, surgical or illness were excluded. The cohort’s inclusion and exclusion criteria are outlined in Supplemental 2.
We removed all patients without a referral for physical therapy entered within 7 days after the index visit, to limit the cohort to only referred patients. Referrals are made electronically within the electronic medical record (EMR) by medical care providers who determine if the referral is needed. Patients are responsible for scheduling the physical therapy appointment.
Outcome variable: physical therapy use
Patient’s use of physical therapy after referral, dichotomized into users and non-users, was our outcome variable. Physical therapy users were identified by the presence of a physical therapy episode of care, within 30 days of referral, to increase the likelihood that their physical therapy was related to their referral. Physical therapists in UUH’s system manually input episodes of care into the EMR each time a patient starts physical therapy.
Factors
Guided by Andersen’s Behavioral Model of Health Services, we grouped factors into predisposing, enabling and need factors that could influence the use of physical therapy (Table 1) (17). Predisposing factors included obesity, age, sex, race and ethnicity. Enabling factors influence a patient’s ability to obtain care and included patients’ distance to the clinic, their payer or health insurance and referring provider. Need factors suggest a patients’ cause of health service use and included patients’ diagnoses and their Charlson Comorbidity Index (CCI) scores. Table 2 describes all factors. Using the patient’s most recent body mass index we determined their obesity status if body mass index was measured 6 months of their index visit. Otherwise, it was considered missing and imputed. We included age as a categorical variable. Distance was determined by using the geocoded location of the patient’s residence and calculating how far they lived, in a straight line, from the closest UUH physical therapy clinic location in kilometres. We categorized distance as 0–10, 10–20 and 20 for interpretability reasons. We included distance to represent the time and resources that a patient will need to get to their physical therapy appointment. Missing distance and body mass index data, both less than 5% of the cohort, were imputed with hotdeck imputation. With hotdeck imputation, the missing values were replaced by random values from the same variable, using the Schonlau implementation from Stata® (18).
Table 1.
Andersen’s Behavior Model of Health
| Predisposing | Enabling | Need |
|---|---|---|
| Obesity | Financing—health plan | Diagnosis |
| Sex | Organization factors: | CCI |
| Age | Distance to clinic | |
| Race | Referring provider | |
| Ethnicity |
Variables in the health record that are hypothesized to be related to a patient’s use of physical therapy.
Table 2.
Comparison of physical therapy users with non-users after referral for a musculoskeletal condition by a medical care provider (1 October 2015 to 30 September 2018)
| Variable | Total | Did not use | Used PT | P-values |
|---|---|---|---|---|
| Full sample, n (%) | 15 870 (100) | 11 065 (76) | 3812 (24) | |
| Obesity (BMI >30) | 5397 (34) | 4275 (35) | 1122 (29) | <0.001 |
| Age, n (%) | 0.007 | |||
| 18–30 | 2837 (18) | 635 (17) | 2202 (18) | |
| 30–40 | 4042 (25) | 1041 (27) | 3001 (25) | |
| a40–50 | 3854 (24) | 949 (25) | 2905 (24) | |
| 50–60 | 3489 (22) | 2686 (22) | 803 (21) | |
| 50–65 | 1655 (10) | 384 (10) | 1271 (11) | |
| Female, n (%) | 8866 (56) | 6750 (56) | 2116 (56) | 0.636 |
| Ethnicity, n (%) | 0.002 | |||
| Not Hispanic/Latino | 13 652 (86) | 10 310 (85) | 3342 (88) | |
| Hispanic/Latino | 1856 (12) | 1470 (12) | 386 (10) | |
| Choose not to disclose | 362 (2) | 281 (2) | 81 (2) | |
| Race, n (%) | 0.022 | |||
| White | 12 692 (80) | 9617 (80) | 3026 (79) | |
| Other | 2349 (15) | 1801 (15) | 548 (14) | |
| Asian | 528 (3) | 372 (3) | 156 (4) | |
| Black or African American | 353 (2) | 271 (2) | 82 (2) | |
| Distance in km, n (%) | <0.001 | |||
| 0–10 | 13 220 (83) | 9713 (81) | 3507 (92) | |
| 10–20 | 1010 (6) | 866 (7) | 144 (4) | |
| 20+ | 1640 (10) | 1482 (12) | 158 (4) | |
| Payer, n (%) | <0.001 | |||
| Commercial | 12 855 (81) | 9636 (80) | 3219 (85) | |
| Medicaid | 1238 (8) | 1002 (8) | 236 (6) | |
| Medicare/government | 571 (4) | 489 (4) | 82 (2) | |
| Self-pay | 724 (5) | 605 (5) | 119 (3) | |
| Workers compensation | 483 (3) | 329 (3) | 154 (4) | |
| Referral department, n (%) | <0.001 | |||
| Orthopaedics | 8161 (51) | 5892 (49) | 2269 (60) | |
| Primary Care | 5092 (32) | 3988 (33) | 1104 (29) | |
| Emergency and Urgent Care | 1268 (8) | 1089 (9) | 179 (5) | |
| Other Specialties | 1349 (8) | 1092 (9) | 257 (7) | |
| Diagnosis code, n (%) | <0.001 | |||
| Lower back pain | 2605 (16) | 2099 (17) | 506 (13) | |
| Cervical | 2041 (13) | 1575 (13) | 466 (12) | |
| Knee | 2599 (16) | 1900 (16) | 699 (18) | |
| Shoulder | 2987 (19) | 2254 (19) | 733 (19) | |
| Hip | 1772 (11) | 1320 (11) | 452 (12) | |
| Elbow | 1160 (7) | 893 (7) | 267 (7) | |
| Ankle | 1038 (7) | 769 (6) | 269 (7) | |
| Wrist/hand | 501 (3) | 382 (3) | 119 (3) | |
| Thoracic | 1034 (7) | 759 (6) | 275 (7) | |
| Arthritis | 133 (1) | 110 (1) | 23 (1) | |
| CCI, n (%) | 0.004 | |||
| 0 | 11 837 (75) | 8928 (74) | 2909 (76) | |
| 1 or greater | 4033 (25) | 3133 (26) | 900 (24) |
BMI, body mass index; km, kilometres.
aIndex factor.
The CCI was dichotomized into low (0) and any (1 or greater) scores. The CCI is calculated in the UUH EDW for every visit using past and current ICD-10 codes. Race included ‘White’, ‘Asian’, ‘Black or African American’ and ‘Other’. Race categories with less than 3% representation were included in ‘Other’ and represented ‘unknown’, ‘Native Hawaiian and other Pacific Islander’, ‘American Indian and Alaska Native’ and ‘Choose Not to Disclose’. Ethnicity was categorized into ‘non-Hispanic/non-Latino’, ‘Hispanic/Latino’ and ‘Choose Not to Disclose/Unknown’. Payers were identified as commercial, Medicaid, Medicare and other government, self-pay and worker’s compensation in the EDW. We grouped the specialization of the referring providers into orthopaedics, primary care, emergency/urgent care and other specialties based on their clinic description. Other specialties included neurology, wellness clinics, obstetrics/gynaecology, occupational medicine, rheumatology, fall clinics, podiatry, employee health, chronic disease clinic, pain management and rheumatology. Patients were grouped using only their primary ICD-10 code from their index visit into lower back pain, cervical, knee, shoulder, hip, elbow, ankle, wrist/hand, thoracic and arthritis diagnosis categories.
Statistical analysis
Descriptive statistics
We compared differences in patient factors between physical therapy users and non-users with chi-square tests for proportions and t-tests for continuous variables. Descriptive statistics for categorical variables are reported with numbers and proportions (%). Descriptive statistics for continuous variables are reported with means and standard deviations (SD). A significance level of P < 0.05 was used for all analyses.
Multivariate regression
Factors were included in our primary regression if their P-value was less than 0.20 in the univariate analyses or if their inclusion resulted in a change of 10% or more in the effect estimate of the primary predictor (19,20). Collinearity diagnostics included examinations of Spearman’s rank correlation coefficients and variance inflation factors. We did not find any significant collinearity or interaction between using a variance inflation factor of <10 and a Spearman’s ρ < 0.20 (21).
To estimate the adjusted measures of association between physical therapy use and patient factors, we used a modified Poisson regression with a robust variance estimator, which provides more interpretable risk ratios (22). We used logistic regression to assess the sensitivity of the model. The sample was split into training and validation samples. We ran a logistic regression using a 75% randomly selected training sample to estimate the model parameters. We then applied the model to the other 25% (validation sample) of the sample to estimate the prediction error for model selection using a receiver operating characteristic (ROC) curve (C-statistic).
STATA® (version 15.1, College Station, TX) was used for all analyses, and the do-file for syntax is included in Supplemental 3.
Results
From 1 October 2015 to 30 September 2018 a total of 90 430 patients were identified with a medical care provider in the UUH system for a musculoskeletal condition (Table 3). The cohort was reduced by 4244 patients after excluding for comorbid conditions and seven patients’ missing diagnosis codes. Of the remaining 86 186 patients, 15 870 (18%) had a referral to physical therapy, and 3812 (24%) of these patients attended physical therapy in the UUH system after referral. Patients who attended physical therapy made up 4% of the total cohort.
Table 3.
Patients who had an index visit with a medical care provider (1 October 2015 to 30 September 2018)
Univariate comparisons of patient factors between users and non-users
Table 2 describes the cohort. Patients referred to physical therapy were on average 43.0 (SD 12.3) years of age, the majority were female (56%), non-Hispanic ethnicity (86%), White race (80%), using commercial insurance (81%) and had a score of 0 on the CCI (75%).
There were differences between users and non-users in referral departments (P < 0.001), ethnicity (P = 0.003), race (P = 0.022), payer (P < 0.001), diagnosis (P < 0.001), CCI (P = 0.004) and distance to a physical therapy clinic (P < 0.001). Users tended to live closer to the physical therapy clinics, identified as not Hispanic/Latino, were white, had no comorbidities, and were referred by orthopaedics. Patients who used physical therapy were less likely to be obese (29%) compared with non-users (35%, P < 0.001). Patients accessing physical therapy treatment were covered by commercial insurance more often (85%) than non-users (80%, P < 0.001).
Predictors of physical therapy use
All factors were statistically, but possibly only marginally, associated with a patient’s use of physical therapy after referral except for sex and CCI scores. Table 4 shows the association between physical therapy use and the included patient and system characteristics. Patient characteristics that reduced the likelihood of physical therapy use were obesity, using Medicare or Medicaid, being referred by primary, or urgent care, age less than 30 years old, and Hispanic/Latino ethnicity. Patients with a diagnosis of knee, ankle, thoracic or hip conditions were more likely to use physical therapy compared with patients with lower back pain diagnoses. The model’s diagnostic validity was poor indicated by the ROC curve of 0.63.
Table 4.
Predictors of physical therapy use after referral for a musculoskeletal condition by a medical care provider (1 October 2015 to 30 September 2018): multivariate Poisson of patient factors
| Variable | Risk ratio (95% CI) | P-value (RR) |
|---|---|---|
| BMI | ||
| Non-obesea | 1.00 (ref) | |
| Obesity | 0.84 (0.79–0.89) | <0.001 |
| Age | ||
| 18–30 | 0.89 (0.81–0.97) | 0.007 |
| 30–40 | 1.03 (0.92–1.11) | 0.489 |
| a40–50 | 1.00 (ref) | |
| 50–60 | 0.95 (0.88–1.03) | 0.206 |
| 50–65 | 0.98 (0.89–1.09) | 0.766 |
| Sex (female) | 1.00 (0.95–1.06) | 0.976 |
| Distance | ||
| 0–10a | 1.00 (ref) | |
| 10–20 | 0.53 (0.46–0.62) | <0.001 |
| 20+ | 0.34 (0.29–0.40) | <0.001 |
| Referral diagnosis code | ||
| Lower back paina | 1.00 (ref) | |
| Cervical | 1.11 (1.00–1.24) | 0.066 |
| Knee | 1.18 (1.07–1.31) | 0.001 |
| Shoulder | 1.12 (1.01–1.24) | 0.025 |
| Hip | 1.18 (1.06–1.32) | 0.003 |
| Elbow | 1.04 (0.91–1.18) | 0.597 |
| Ankle | 1.16 (1.02–1.32) | 0.021 |
| Wrist/hand | 1.11 (0.93–1.31) | 0.257 |
| Thoracic | 1.27 (1.12–1.45) | <0.001 |
| Arthritis | 0.90 (0.62–1.28) | 0.549 |
| Race | ||
| aWhite | 1.00 (ref) | |
| Other | 1.15 (1.04–1.27) | 0.007 |
| Asian | 1.11 (0.97–1.28) | 0.122 |
| Black or African American | 1.05 (0.87–1.29) | 0.572 |
| Ethnicity | ||
| aNot Hispanic/Latino | 1.00 (ref) | |
| Hispanic/Latino | 0.81 (0.73–0.92) | <0.001 |
| Choose not to disclose | 0.83 (0.68–1.02) | 0.073 |
| Payer | ||
| aCommercial | 1.00 (ref) | |
| Medicaid | 0.83 (0.74–0.94) | 0.003 |
| Medicare and other government | 0.67 (0.54–0.81) | <0.001 |
| Self-pay | 0.72 (0.61–0.86) | <0.001 |
| Workers compensation | 1.50 (1.31–1.72) | <0.001 |
| CCI | ||
| a0 | 1.00 (ref) | |
| 1 or greater | 0.96 (0.90–1.03) | 0.290 |
| Referral department | ||
| aOrthopaedics | 1.00 (ref) | |
| Primary Care | 0.77 (0.72–0.82) | <0.001 |
| Emergency and Urgent Care | 0.53 (0.46–0.61) | <0.001 |
| Specialty | 0.71 (0.63–0.79) | <0.001 |
BMI, body mass index; CI, confidence interval; km, kilometres; RR, risk ratio.
aIndex factor.
Conclusions
Our retrospective cohort study is the first to describe patients’ rates of physical therapy use in a broad population of musculoskeletal conditions after a referral. A strength of this study is the use of Andersen’s Behavioral Model of Health Services Use to guide model development and analysis. Past literature separately examined rates of referral and physical therapy use rates and was often limited to specific diagnostic groups (9,12,16,23–25). The low rates of referral and use of physical therapy in our cohort are consistent with past literature conducted in one acute low back pain population, two chronic spine pain populations, and in general musculoskeletal populations (7,15,23,26,27). While a low rate in a population with a chronic disease may be understandable, the low rate of physical therapy use in patients with new or exacerbated symptoms was surprising. Based on clinical practice guidelines advocating for physical therapy treatments and high medical care provider’s referral rates, we expected better uptake (1,5,11).
Our analyses suggest that patient choice and access are significant drivers of physical therapy use. In Andersen’s model, health care use is attributed to three characteristics: predisposing, enabling and need factors (28). We included predisposing factors primarily due to their availability in the EDW with limited representation of enabling and need factors. Enabling and need factors influenced patients’ physical therapy use, similar to other studies (29). These factors must be explored further by health systems, clinicians, policymakers and researchers to address low referral and use rates.
Predisposing factors are demographic and social factors at the patient level that can be associated with health behaviours. Unlike past studies in other health systems that show female and older patients tend to use more health services often in non-musculoskeletal diagnoses, we found that in patients over 30 years old, age and sex were not related to physical therapy use (30,31). We also found that obesity and Hispanic/Latino ethnicity were related to a reduced likelihood of using physical therapy as expected based on past research (29,32). Patients who are obese or of Hispanic/Latino ethnicity may experience greater barriers to using physical therapy compared with non-Hispanic and non-obese patients, including limited resources, anxiety/depression, social support and documentation status (33,34).
Need factors represent a patient’s perceived and actual need for health care. In this analysis, we included comorbidities and diagnoses as indicators of a patient’s need for physical therapy use. A higher CCI was not predictive of increased physical therapy use contrary to prior research showing that higher CCIs represent an increased need for health services (35,36). Lower back pain diagnoses were associated with less physical therapy use compared with knee, hip, ankle and thoracic diagnoses, which was unexpected since there is an abundance of research supporting the use of physical therapy for lower back pain (7,12,37).
Enabling factors consider the influence of patient’s resources and their access to care. Our results suggest that enabling factors, including commercial insurance, closer distance to a clinic, and being referred by orthopaedics may increase a patient’s likelihood of accessing physical therapy. Relative to orthopaedic providers, other providers may be more focussed on a patient’s medical stability and not their musculoskeletal condition. An orthopaedic provider may provide patients with a clear diagnosis and treatment plan that includes physical therapy. The referral may be discussed in greater detail and a relationship may exist between the orthopaedic and physical therapy clinics that improves access to physical therapy. Patients who lived further from a UUH physical therapy clinic had a reduced likelihood of using physical therapy, as expected since travelling further distances likely increases patient burden. Health systems should consider the influence of distance, referral source and payer type as important factors in patients’ ability to access physical therapy.
This study should be viewed in light of several limitations, including those familiar to retrospective design relying on administrative data from one health system in the USA. Although our sample was similar to past research, it was not representative of the general US population (27,38). Essential factors such as psychological variables, social determinants of health, etc. are not available within the UUH EMR limiting our analyses. Several enabling and need factors, including acuity, patient perceptions and patient burden (e.g. co-pays, wait times, and clinic hours) towards accessing physical therapy, were not represented in our model likely contributing to the modest model fit and limiting our relevance to other settings (39,40). Furthermore, we could only include care provided within the UUH system and did not include any secondary or tertiary diagnosis codes. Some patients classified as non-users of physical therapy may have received physical therapy care in a non-UUH clinic and our wash-out period may have misclassified some patients as having a new musculoskeletal condition.
Our study is the first to present associations between patient factors, available from EMR records, and patients’ use of physical therapy after referral for musculoskeletal conditions. Low levels of use after referral suggest that many patients who may benefit from seeing a physical therapist are not accessing those services. Researchers and learning health systems should consider our low rates of physical therapy use after referral and our findings when initiating efforts to improve musculoskeletal care. Providers may be recommending optimal care, but if a patient does not adhere to this recommendation, outcomes may be limited. Improving access will require improved methods for understanding decision-making at the patient level and how to support patients in adhering to best care models. Health systems and providers may be able to increase the use of physical therapy by supporting patients when navigating the health system by reducing barriers (41). The poor predictive ability of our model and limited external validity underscore the need for more detailed investigations into the factors that influence a patient’s choice or ability to access physical therapy care after a referral.
Further work is needed to understand the influence that physical therapy use has on outcomes and other medical service use for a variety of musculoskeletal conditions. Our analysis was unable to explore how patients make choices in the health care system. Qualitative research is another necessary next step to explore the influence of costs, pain levels, acuity of symptoms, provider recommendations and other potential barriers to accessing care.
Supplementary Material
Declaration
Funding: this research was funded in part by the Promotion of Doctoral Studies I and II Scholarships from the Foundation for Physical Therapy Research.
Ethical approval: the study has been approved by the Institutional Review Board at the University of Utah.
Conflict of interest: None declared besides the support from the Foundation of Physical Therapy Research.
Clinical trials registration
Not applicable.
References
- 1. Lewis R, Gómez Álvarez CB, Rayman Met al.. Strategies for optimising musculoskeletal health in the 21st century. BMC Musculoskelet Disord 2019; 20 (1): 164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Sinnott PL, Dally SK, Trafton J, Goulet JL, Wagner TH. Trends in diagnosis of painful neck and back conditions, 2002 to 2011. Medicine 2017; 96 (20): e6691. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Vos T, Abajobir AA, Abbafati Cet al.. Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390 (10100): 1211–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Masaracchio M, Hanney WJ, Liu X, Kolber M, Kirker K. Timing of rehabilitation on length of stay and cost in patients with hip or knee joint arthroplasty: a systematic review with meta-analysis. PLoS One 2017; 12 (6): e0178295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Qaseem A, Wilt TJ, McLean RM, Forciea MA. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians. Ann Intern Med 2017; 166 (7): 514–30. [DOI] [PubMed] [Google Scholar]
- 6. Kazis LE, Ameli O, Rothendler Jet al. Observational retrospective study of the association of initial healthcare provider for new-onset low back pain with early and long-term opioid use. BMJ Open 2019; 9 (9): e028633. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Frogner BK, Harwood K, Andrilla CHA, Schwartz M, Pines JM. Physical therapy as the first point of care to treat low back pain: an instrumental variables approach to estimate impact on opioid prescription, health care utilization, and costs. Health Serv Res 2018; 53 (6): 4629–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Horn ME, Fritz JM. Timing of physical therapy consultation on 1-year healthcare utilization and costs in patients seeking care for neck pain: a retrospective cohort. BMC Health Serv Res 2018; 18 (1): 887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Bornhöft L, Thorn J, Svensson Met al.. More cost-effective management of patients with musculoskeletal disorders in primary care after direct triaging to physiotherapists for initial assessment compared to initial general practitioner assessment. BMC Musculoskelet Disord 2019; 20 (1): 186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Arnold E, La Barrie J, DaSilva Let al.. The effect of timing of physical therapy for acute low back pain on health services utilization: a systematic review. Arch Phys Med Rehabil 2019; 100 (7): 1324–38. [DOI] [PubMed] [Google Scholar]
- 11. Chou R, Deyo R, Friedly Jet al. Nonpharmacologic therapies for low back pain: systematic review for an American College of Physicians Clinical Practice Guideline. Ann Intern Med 2017; 166 (7): 493–505. [DOI] [PubMed] [Google Scholar]
- 12. Thackeray A, Hess R, Dorius J, Brodke D, Fritz J. Relationship of opioid prescriptions to physical therapy referral and participation for Medicaid patients with new-onset low back pain. J Am Board Fam Med 2017; 30 (6): 784–94. [DOI] [PubMed] [Google Scholar]
- 13. Zheng P, Kao MC, Karayannis NV, Smuck M. Stagnant physical therapy referral rates alongside rising opioid prescription rates in patients with low back pain in the United States 1997–2010. Spine (Phila Pa 1976) 2017; 42 (9): 670–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Dennis S, Watts I, Pan Y, Britt H. The likelihood of general practitioners referring patients to physiotherapists is low for some health problems: secondary analysis of the Bettering the Evaluation and Care of Health (BEACH) observational study. J Physiother 2018; 64 (3): 178–82. [DOI] [PubMed] [Google Scholar]
- 15. Freburger JK, Khoja S, Carey TS. Primary care physician referral to physical therapy for musculoskeletal conditions, 2003–2014. J Gen Intern Med 2018; 33 (6): 801–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Aguirre F, Heft J, Yunker A. Factors associated with nonadherence to pelvic floor physical therapy referral for the treatment of pelvic pain in women. Phys Ther 2019; 99 (7): 946–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Babitsch B, Gohl D, von Lengerke T. Re-revisiting Andersen’s Behavioral Model of Health Services Use: a systematic review of studies from 1998–2011. Psychosoc Med 2012; 9:Doc11. doi: 10.3205/psm000089 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Schonlau M. HOTDECKVAR: Stata module for hotdeck imputation, Statistical Software Components S458527. Boston College Department of Economics. 2018. [Google Scholar]
- 19. Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993; 138 (11): 923–36. [DOI] [PubMed] [Google Scholar]
- 20. Thompson WW, Price C, Goodson Bet al. ; Vaccine Safety Datalink Team. Early thimerosal exposure and neuropsychological outcomes at 7 to 10 years. N Engl J Med 2007; 357 (13): 1281–92. [DOI] [PubMed] [Google Scholar]
- 21. Yoo W, Mayberry R, Bae Set al.. A study of effects of multicollinearity in the multivariable analysis. Int J Appl Sci Technol 2014; 4 (5): 9–19. [PMC free article] [PubMed] [Google Scholar]
- 22. Knol MJ. Down with odds ratios: risk ratios in cohort studies and randomised clinical trials. Ned Tijdschr Geneeskd 2012; 156 (28): A4775. PMID: 22805792. [PubMed] [Google Scholar]
- 23. Bath B, Lawson J, Ma D, Trask C. Self-reported use of family physician, chiropractor and physiotherapy services among adult Canadians with chronic back disorders: an observational study. BMC Health Serv Res 2018; 18 (1): 970. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Postler A, Luque Ramos A, Goronzy Jet al.. Prevalence and treatment of hip and knee osteoarthritis in people aged 60 years or older in Germany: an analysis based on health insurance claims data. Clin Interv Aging 2018; 13: 2339–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Lange T, Luque Ramos A, Albrecht Ket al.. Prescription frequency of physical therapy and analgesics before total hip and knee arthroplasy: an epidemiological analysis of routine health care data from Germany. Orthopade 2018; 47 (12): 1018–26. [DOI] [PubMed] [Google Scholar]
- 26. Callhoff J, Albrecht K, Hoffmann Fet al.. Reality of care for musculoskeletal diseases at the population level: results of the PROCLAIR collaborative project. Z Rheumatol 2019; 78 (suppl 2): 73–9. [DOI] [PubMed] [Google Scholar]
- 27. Freburger JK, Holmes GM, Carey TS. Physician referrals to physical therapy for the treatment of musculoskeletal conditions. Arch Phys Med Rehabil 2003; 84 (12): 1839–49. [DOI] [PubMed] [Google Scholar]
- 28. Andersen RM. National health surveys and the behavioral model of health services use. Med Care 2008; 46 (7): 647–53. [DOI] [PubMed] [Google Scholar]
- 29. Mbada C, Olawuyi A, Oyewole OOet al.. Characteristics and determinants of community physiotherapy utilization and supply. BMC Health Serv Res 2019; 19 (1): 168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Dieleman JL, Squires E, Bui ALet al.. Factors associated with increases in US health care spending, 1996–2013. JAMA 2017; 318 (17): 1668–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Keil J, Brendler V, Sachse Cet al.. Gender-specific differences in the utilization of health care services in an urban population sample. Gesundheitswesen 2019; 82 (3): e17–23. doi: 10.1055/a-0820-3584 [DOI] [PubMed] [Google Scholar]
- 32. Rommel A, Kroll LE. Individual and regional determinants for physical therapy utilization in Germany: multilevel analysis of national survey data. Phys Ther 2017; 97 (5): 512–23. [DOI] [PubMed] [Google Scholar]
- 33. Vader K, Doulas T, Patel R, Miller J. Experiences, barriers, and facilitators to participating in physical activity and exercise in adults living with chronic pain: a qualitative study. Disabil Rehabil 2019: 1–9. [DOI] [PubMed] [Google Scholar]
- 34. Flannery C, McHugh S, Anaba AEet al. Enablers and barriers to physical activity in overweight and obese pregnant women: an analysis informed by the theoretical domains framework and COM-B model. BMC Pregnancy Childbirth 2018; 18 (1): 178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Bottle A, Honeyford K, Chowdhury F, Bell D, Aylin P. Factors associated with hospital emergency readmission and mortality rates in patients with heart failure or chronic obstructive pulmonary disease: a national observational study. Health Serv Deliv Res 2018; 6 (26). doi: 10.3310/hsdr06260 [DOI] [PubMed] [Google Scholar]
- 36. García-Olmos L, Aguilar R, Lora Det al. ; CHIC Group. Development of a predictive model of hospitalization in primary care patients with heart failure. PLoS One 2019; 14 (8): e0221434. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Horn ME, George SZ, Fritz JM. Influence of initial provider on health care utilization in patients seeking care for neck pain. Mayo Clin Proc Innov Qual Outcomes 2017; 1 (3): 226–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. U.S. Census Bureau QuickFacts . US Census Bureau. https://www.census.gov/quickfacts/fact/table/US/PST045219. 9 July 2019 (accessed on 20 August 2020). [Google Scholar]
- 39. Kirby ER, Broom AF, Adams J, Sibbritt DW, Refshauge KM. A qualitative study of influences on older women’s practitioner choices for back pain care. BMC Health Serv Res 2014; 14: 131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Joelsson M, Bernhardsson S, Larsson ME. Patients with chronic pain may need extra support when prescribed physical activity in primary care: a qualitative study. Scand J Prim Health Care 2017; 35 (1): 64–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Bhatt J, Bathija P. Ensuring access to quality health care in vulnerable communities. Acad Med 2018; 93 (9): 1271–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.