Assessing the Prioritization of Primary Care Referrals for Polysomnograms

Abstract

Study Objective:

The mortality attributed to obstructive sleep apnea (OSA) is comparable to that of breast cancer and colon cancer. We sought to determine if patients at high risk for OSA were less likely to be referred by their primary care physician for polysomnograms (PSG) than mammograms or endoscopies.

Design:

Prospective cohort study; patients were recruited between January 2007 and April 2007.

Setting:

Academic public hospital system

Patients:

395 patients waiting for family or internal medicine primary care appointments were administered the Berlin questionnaire. Chart abstraction or interview determined demographics; insurance and employment status; body mass index (BMI); comorbidities; and prior PSG, mammography, or endoscopy referrals.

Results:

Mean BMI was 30 ± 7.4 kg/m²; 187 (47%) patients had high-risk Berlin scores. Overall, 19% of patients with high-risk Berlin scores were referred for PSG, compared to 63% of those eligible for mammograms and 80% of those eligible for endoscopies. Women (OR = 2.9, P = 0.02), COPD (OR = 4.6, P = 0.03), high-risk Berlin scores (OR = 3.4, P = 0.009), and higher BMI (OR = 1.1, P < 0.001) were positively associated with PSG referrals. Privately insured patients were less likely to be referred than uninsured patients (OR = 0.3, P = 0.04). There was no significant difference in referrals among those with other forms of insurance. Race was not associated with PSG referrals.

Conclusion:

In a public hospital, primary care patients were less likely to be referred for PSG compared to mammogram and endoscopy. Uninsured patients were more likely to be referred for PSG than those with private insurance. Further studies are needed to address the low PSG referral rates in high-risk populations.

Citation:

Thornton JD; Chandriani K; Thornton JG; Farooq S; Moallem M; Krishnan V; Auckley D. Assessing the prioritization of primary care referrals for polysomnograms. SLEEP 2010;33(9):1255-1260.

EVIDENCE CONTINUES TO EMERGE IDENTIFYING THE NUMEROUS HEALTH EFFECTS ASSOCIATED WITH UNTREATED OBSTRUCTIVE SLEEP APNEA (OSA).^1–7 With recent epidemiologic studies suggesting an at-risk prevalence as high as 26% of the US adult population,^8,9 and a mortality as high as 20 deaths per 1,000 person-years,¹⁰ the nationwide impact of untreated OSA is significant.¹¹ Fortunately, continuous positive airway pressure (CPAP) is a cost-effective treatment for the majority of moderately affected patients.^12,13 However, proper OSA treatment requires appropriate and expensive diagnosis in the form of polysomnography (PSG).¹⁴

Primary care physicians have a unique opportunity to intervene in this disease, given the high percentage of symptomatic patients presenting to their practices.^15–20 However, studies suggest that primary care physicians may not be aware of the adverse outcomes associated with untreated OSA.¹⁷ As a result, they may be less likely to address OSA during health maintenance visits.^21,22

We sought to examine the frequency of and factors associated with polysomnogram referrals for patients at high risk for OSA attending primary care clinics at an inner-city public hospital. Because of the expense associated with polysomnography, we were interested in the role insurance status played in polysomnogram referrals. We were also interested in knowing if such patients underwent referral for polysomnograms at the same rate as they underwent referral for 2 other common screening tests: mammograms for breast cancer, and fecal occult blood testing (FOBT), colonoscopies, or flexible sigmoidoscopies for colon cancer. Breast and colon cancer were chosen as comparative diseases because their mortality rates are similar to that of OSA.^10,23

METHODS

Participants were recruited between January and April 2007 from 2 general internal medicine clinics and one family medicine clinic affiliated with an urban safety-net hospital. One general internal medicine clinic was at a site away from the main hospital campus, while the other 2 clinics were on site. The clinics have a combined patient volume of over 2,500 visits per month. Two study investigators visited each clinic up to 3 h daily and directly approached 461 patients while they waited to visit with their primary care physician. If patients had time to complete the interview, indicated interest, met enrollment criteria, and provided consent, the interviews were conducted in a private setting. Patients were enrolled if they spoke English, were older than 18 years of age, were able to be interviewed, and if they had been seen by their primary caregiver at least 3 times in the last 2 years. The MetroHealth Medical Center Institutional Review Board approved the study.

Data Collection

The Berlin questionnaire was administered to all participants. The Berlin questionnaire was designed to screen for sleep apnea in a primary care population and stratifies patients into those at low or high risk. It has been validated, has high reliability (Cronbach α 0.86–0.92), and has high positive predictive value in the ambulatory setting for identifying patients at high risk of sleep apnea.²⁴ The Berlin questionnaire has been used by several studies as a surrogate means to determine OSA prevalence,^8,25 and may be superior to the Epworth Sleepiness Scale in identifying those at risk for complications of OSA.²⁶ Patients were also asked to provide their demographic information, including race, ethnicity, and employment status.

Two investigators (SF and JGT) who were blinded to the interview and questionnaire results queried the patients' electronic medical records from their enrollment in the health care system until the date of their interview to determine which patients had been referred for a polysomnogram or had an existing diagnosis of sleep apnea. All patients had at least 2 years of medical records reviewed. The medical records were further examined for the presence of an existing referral for a mammogram (or an existing diagnosis of breast cancer), flexible sigmoidoscopy, or colonoscopy (or an existing diagnosis of colon cancer) using eligibility criteria from the US Preventive Services Task Force.^27,28 The US Preventive Services Task Force considers FOBT to be a reasonable alternative to colonoscopy.²⁸ We included the number of patients who underwent FOBT in those who underwent colon cancer screening. Because many primary care physicians may refer patients to subspecialty clinics in lieu of directly ordering screening procedures such as colonoscopy, we considered referrals to gastroenterology clinic the same as colon cancer screening and referrals to sleep clinic the same as OSA screening. The focus of this study was on physician behavior and not patient behavior. As a result, we did not examine whether patients actually underwent the procedures that were ordered. The body mass index (BMI) and comorbidities of each patient were also obtained from the electronic medical record. Comorbidities that have been previously shown to be strongly associated with obstructive sleep apnea were chosen a priori for the analysis. Insurance status was obtained from each patient's medical record at the time of the interviews. Insurance status was exclusively categorized into “uninsured,” “Medicaid,” “Medicare,” or “private or commercial.” Those with multiple payers were included in the group with the higher reimbursement rates.

Following completion of the project in 2007, we sent letters to all of the physicians of patients in the study identified to be at high-risk for OSA. Two years later, we assessed whether those patients were subsequently referred.

Statistical Analysis

The primary outcome measure was referral for a polysomnogram or referral to sleep clinic. Additional outcome measures were referral for colonoscopy, flexible sigmoidoscopy, or to gastroenterology clinic; order for fecal occult blood testing; and referral for mammogram. Student t-test or analysis of variance (ANOVA) was used to compare the descriptive statistics of patients. Multivariate logistic regression was used to determine the factors that were each associated with polysomnogram referral. These factors were defined a priori and included Berlin score, body mass index (BMI), insurance status, employment status, gender, age, clinic type, each comorbidity, race, and ethnicity. Pairwise correlations were performed between those referred for screening for sleep apnea, breast cancer, or colon cancer to quantify the relationships between each. A 2-sided P-value < 0.05 was considered statistically significant. Analyses were performed using STATA Statistical Software (Release 9.0, Stata Corportation, College Station, TX).

RESULTS

The characteristics of the 395 patients who were enrolled in the study (86% of those approached) are found in Table 1. They were cared for by 120 physicians. High-risk Berlin scores were found in 187 (47%) patients. Those with high-risk Berlin scores were more likely to be obese (BMI 33 vs. 28, P < 0.001) and to have comorbid conditions (14 vs. 13, P < 0.001). Hypertension, diabetes, hyperlipidemia, coronary artery disease, depression, and asthma were significantly more common in patients with high-risk Berlin scores. Race, ethnicity, and gender were not associated with having a high-risk Berlin score. Patients with high-risk Berlin scores were more likely to be referred for a polysomnogram (19% vs. 4%, P < 0.001), but the number of patients referred for polysomnogram in the entire cohort was very low (44 [11%]).

Table 1.

Patient characteristics

	Low-Risk Berlin	High-Risk Berlin
Characteristics	N = 208 (%)	N = 187 (%)	P-value
Age (mean ± SD)	47 ± 15	49 ± 13	0.2
Women	124 (60%)	119 (64%)	0.4
Insurance
No Insurance	50 (24%)	56 (30%)
Medicaid	48 (23%)	43 (23%)	0.5
Medicare	37 (18%)	27 (14%)
Private	73 (35%)	61 (33%)
Race and ethnicity
White	117 (56%)	111 (59%)
African American	62 (30%)	60 (32%)	0.4
Hispanic	19 (9%)	12 (7%)
Other	10 (5%)	4 (2%)
Clinic type
Family medicine	66 (32%)	54 (29%)
Internal medicine satellite	37 (18%)	42 (22%)	0.5
Internal medicine main campus	105 (50%)	91 (49%)
BMI (mean ± SD)	28 ± 6.7	33 ± 7.2	< 0.001
Number of comorbid conditions (mean ± SD)	13 ± 1.4	14 ± 1.6	< 0.001
Comorbid conditions
Hypertension	61 (29%)	101 (54%)	< 0.001
Diabetes	31 (15%)	50 (27%)	0.004
Hyperlipidemia	51 (24%)	73 (39%)	0.002
CHF	5 (2%)	12 (6%)	0.05
Prior MI	4 (2%)	11 (6%)	0.04
CVA	5 (2%)	9 (5%)	0.2
Depression	45 (22%)	65 (35%)	0.004
Bipolar disorder	14 (7%)	11 (6%)	0.7
Thyroid disorder	9 (4%)	17 (9%)	0.06
COPD	11 (5%)	10 (5%)	1.0
Asthma	18 (9%)	36 (19%)	0.002
Chronic kidney disease	1 (0.5%)	4 (2%)	0.1

Patient characteristics stratified by polysomnogram referral are found in Table 2. Table 3 lists those factors associated with referral for polysomnogram in unadjusted and adjusted analyses. Factors significantly associated with polysomnogram referral were women (OR = 2.9, 1.2-7.0), having a diagnosis of chronic obstructive pulmonary disease (COPD) (OR = 4.6, 1.2-18.4), having a positive Berlin score (OR = 3.4, 1.3-8.4), and elevated BMI (OR = 1.1, 1.1-1.2). Compared to those who were uninsured, those with private insurance had a 70% lower odds of being referred for polysomnogram (OR = 0.3, 0.1-0.9). There was no significant difference between those with Medicaid or Medicare and those who were uninsured in referral for polysomnogram.

Table 2.

Patient characteristics by polysomnogram referral

	Not Referred for Poly-somnogram	Referred for Poly-somnogram
Characteristics	N = 351 (%)	N = 44 (%)	P-value
Age (mean ± SD)	48 ± 15	50 ± 11	0.3
Women	209 (60%)	119 (77%)	0.02
Insurance
No Insurance	91 (26%)	15 (34%)
Medicaid	82 (23%)	9 (20.5%)	0.4
Medicare	55 (16%)	9 (20.5%)
Private	123 (35%)	11 (25%)
Race and ethnicity
White	200 (57%)	28 (64%)
African American	111 (31%)	11 (25%)	0.8
Hispanic	27 (8%)	4 (9%)
Other	13 (4%)	1 (2%)
Clinic type
Family medicine	107 (30%)	13 (30%)
Internal medicine satellite	72 (21%)	7 (16%)	0.7
Internal medicine main campus	172 (49%)	24 (54%)
BMI (mean ± SD)	29 ± 6.7	38 ± 8.8	< 0.001
Number of comorbid conditions (mean ± SD)	14 ± 1.5	14 ± 1.4	0.02
Comorbid conditions
Hypertension	133 (38%)	29 (66%)	< 0.001
Diabetes	72 (20%)	9 (20%)	1.0
Hyperlipidemia	106 (30%)	18 (41%)	0.1
CHF	13 (4%)	4 (9%)	0.1
Prior MI	13 (4%)	2 (5%)	0.8
CVA	12 (3%)	2 (5%)	0.7
Depression	97 (28%)	13 (29%)	0.8
Bipolar disorder	24 (7%)	1 (2%)	0.2
Thyroid disorder	22 (6%)	4 (9%)	0.5
COPD	15 (4%)	6 (14%)	0.009
Asthma	46 (13%)	8 (18%)	0.3
Chronic kidney disease	5 (1%)	0	0.4

Table 3.

Factors associated with referral for polysomnogram

Predictor	Unadjusted Odds Ratio (95% CI)	P-value	Adjusted Odds Ratio (95% CI)	P-value
Insurance (None = referent)
Medicaid	0.7 (0.3 – 1.6)	0.4	0.4 (0.1 – 1.3)	0.1
Medicare	1.0 (0.4 – 2.4)	1.0	0.9 (0.3 – 3.0)	0.9
Private	0.5 (0.2 – 1.2)	0.1	0.3 (0.1 – 0.9)	0.04
Employment Status (Unemployed = referent)
Part Time	0.4 (0.1 – 1.9)	0.3	0.3 (0.1 – 1.8)	0.2
Full Time	0.9 (0.5 – 1.9)	0.9	2.1 (0.7 – 6.0)	0.2
Race and ethnicity (white = referent)
African American	0.7 (0.3 – 1.5)	0.3	0.5 (0.2 – 1.3)	0.2
Hispanic	1.0 (0.3 – 3.2)	0.9	1.5 (0.4 – 5.7)	0.6
Women	2.3 (1.1 – 4.8)	0.03	2.9 (1.2 – 7.0)	0.02
COPD	3.5 (1.3 – 9.7)	0.01	4.6 (1.2 – 18.4)	0.03
High-Risk Berlin Score	5.9 (2.7 – 13)	< 0.001	3.4 (1.3 – 8.4)	0.009
BMI (+1 kg/m2)	1.2 (1.1 – 1.2)	< 0.001	1.1 (1.1 – 1.2)	< 0.001
Clinic type (family medicine = referent)
Internal medicine satellite	0.8 (0.3 – 2.1)	0.6	0.4 (0.1 – 1.4)	0.2
Internal medicine main site	1.1 (0.6 – 2.3)	0.7	1.2 (0.5 – 3.0)	0.6

Overall, 105 of the 166 eligible persons (63%) were referred for mammograms, and 138 of the 173 (80%) eligible persons were referred for colonoscopies, flexible sigmoidoscopies, or fecal occult blood tests, compared to 36 of the 187 (19%) eligible persons for polysomnograms (Figure 1A). When limiting the eligible persons to women, 28 of 119 (23%) underwent polysomnograms, compared to 63% for mammograms and 81% for endoscopies or fecal occult blood tests (Figure 1B). Among all 3 referrals for screening, PSG had the highest percentage of uninsured patients (Table 4). However, among those who underwent colorectal cancer screening, fecal occult blood testing had a higher percentage of uninsured than endoscopy. Overall, there was a poor correlation between those patients who were referred for polysomnograms and those who were referred for mammograms (−0.09), colonoscopies (0.06), and fecal occult blood tests (−0.06).

Figure 1A.

The percentage of all eligible female patients referred for obstructive sleep apnea, breast cancer, and colon cancer screening

Table 4.

Relationship between insurance status and sleep apnea, breast cancer, and colon cancer screening

Insurance Type	PSG (n = 39)	Mammogram (n = 101)	Endoscopy or Fecal Occult Blood Testing (n = 133)	Endoscopy (Colonoscopy or Flexible Sigmoidoscopy) (n = 87)	Fecal Occult Blood Testing (n = 51)
No Insurance	13 (33%)	22 (22%)	35 (26%)	19 (22%)	17 (33%)
Medicaid	7 (18%)	16 (16%)	24 (18%)	14 (16%)	10 (20%)
Medicare	8 (21%)	35 (34%)	39 (30%)	31 (36%)	11 (22%)
Private	11 (28%)	28 (28%)	35 (26%)	23 (26%)	13 (25%)

Figure 1B.

The percentage of all eligible patients referred for obstructive sleep apnea, breast cancer, and colon cancer screening

Two-Year Follow-Up

Two years following conclusion of the study, 86% of those with positive Berlin scores who had not been referred for a polysomnogram by the end of the study still had not been referred. This occurred despite the study authors notifying the primary care physicians that these patients were at high risk for OSA.

DISCUSSION

In this study, we found that primary care physicians referred their patients for OSA screening significantly less frequently than for colorectal and breast cancer screening. However, uninsured patients were more likely to be referred for OSA screening than privately insured persons with similar severity of illness, socioeconomic status, and risk for obstructive sleep apnea. This suggests that primary care physicians are not using lack of health insurance as a barrier in OSA screening referrals, but they also are not prioritizing OSA screening as highly as screening for cancer. Recent data suggests that the mortality rate due to obstructive sleep apnea (20.3 deaths/1,000 person-years) is comparable to that of breast cancer (24.5 deaths/100,000 women-years) and colon cancer (18.2 deaths/100,000 person-years).^10,23 If accurate, this should be a point of emphasis in campaigns directed at educating those interested in decreasing the adverse events from this disease.

The referral rate for polysomnograms of patients in this study with a high-risk Berlin score was 19%, while the referral rates among eligible patients for mammograms and endoscopies or FOBT were 63% and 80%, respectively. Procedural costs are not likely to be the primary reason, as PSG and colonoscopy are similar in cost.^14,29 The substantial difference in PSG referral rate compared to that for mammograms and colonoscopies may be explained by the extensive awareness-raising campaigns by medical organizations,³⁰ community-based health organizations, fundraising organizations,³¹ and celebrities promoting screening for breast and colon cancer.^32,33 Physician incentives have also been touted as a means to increase cancer screening.³⁰

Several additional reasons could account for the low PSG referral rate in patients who are at high risk for obstructive sleep apnea. During primary care appointments, physicians may spend a portion of the encounter discussing health-promoting behaviors such as avoiding smoking, exercising, eating a balanced diet, and wearing seat belts, but may fail to address receipt of appropriate sleep as a form of health promotion. In a study surveying primary care physicians about their attitudes regarding health promotion, there was no mention of discussing a patient's sleep.²² Similarly, among a group of medical interns who had not specifically received instruction about sleep medicine, only 13% asked questions about a patient's sleep patterns.³⁴ If sleep is not routinely discussed as part of a patient's history or during a conversation about health-promoting behaviors, irregularities in sleep may also be less likely to be addressed.

Sleep apnea is a relatively new diagnosis, first described in medical literature in 1965.³⁵ As a result, primary care physicians may still be less aware of symptoms, risk factors, diagnosis, and consequences of this disorder. A recent study found that although primary care physicians were aware of the process for diagnosing sleep apnea, in practice they did not identify the patients that needed diagnostic testing nor did they discuss some of the relevant and risky consequences with their patients.²¹ This may also be due to growing administrative requirements and time constraints, making it challenging for physicians to prioritize sleep apnea in a routine care visit. Additionally, doctors may not be receiving adequate training in medical school and residency to appropriately identify patients at risk for sleep apnea.³⁶

The physicians of the patients in this study were sent letters following the completion of the project notifying them of which of their patients had high-risk Berlin scores that were considered high risk for OSA. Two years later, 86% of those with a high-risk Berlin scores who had not been referred by the end of the study still had not been referred for a sleep study. This suggests that even with reminders, primary care physicians may be reticent to screen patients for OSA.

We had hypothesized that a patient's insurance status would influence polysomnogram referral, given the important role insurance has played in influencing physician behavior in other segments of healthcare. Prior studies have shown that uninsured patients receive less care,³⁷ are more likely to visit the emergency room,³⁸ have more advanced disease at the time of diagnosis,³⁹ and have a higher risk of all-cause mortality⁴⁰ compared to insured patients. Unexpectedly, our study showed that uninsured patients are more likely to be referred for PSG than privately insured patients, while there was no significant difference in referral rates when comparing the uninsured to patients with Medicare or Medicaid. One possible explanation for this finding is that primary care physicians in this cohort saw a greater need to refer uninsured patients for polysomnograms compared to privately insured patients. The inner-city public hospital in our study serves a significant portion of the poor and uninsured for Cuyahoga County. Doctors in this hospital system are experienced in dealing with this patient population and therefore may be more in tune with the unique needs of their uninsured patients. Alternatively, other unmeasured factors may be confounding the relationship between insurance status and PSG referral that were not adequately adjusted for in our analysis. Additional work is needed to better understand the factors responsible for our findings.

We found a significant positive association between polysomnogram referral and a diagnosis of COPD. To our knowledge, this is the first time this association has been described. This finding may be related to a heightened awareness among primary care providers of the association and adverse effects of OSA and COPD.⁴¹ Alternatively, primary care physicians could be referring patients with COPD to pulmonologists who are in turn referring the patients for PSG. It is interesting that a similar association was not found between asthma and referrals for PSG. This finding deserves further exploration.

There are limitations to this study. The population of patients in this study was confined to a single, urban public medical system. Despite this focus, patients were recruited from three different internal medicine or family medicine clinics, of which two were located on the main hospital campus and the third at a satellite center. The choice of multiple clinics both on and off site generated a medically, culturally, geographically, and economically diverse group of participants. Additionally, each site was visited multiple times, thereby reducing the risk of sampling bias. Involvement in the study required that each patient saw the same primary care physician in at least three prior visits in the last two years to ensure that physicians had ample time to assess risk factors for OSA and refer patients for polysomnogram. Each patient's electronic medical record was examined comprehensively for past diagnoses and referrals, including those made at outside facilities as documented in the primary care physician's note. While the results of this study may be generalizable to patients at other urban public hospitals, they are by no means representative of patients in other medical systems. Similar studies should be conducted in other settings to note if similar findings occur in broader populations.

The severe adverse consequences of untreated OSA range from cardiovascular disease to motor vehicle accidents. Despite these significant complications, the referral rate of patients remains low, and those who are referred tend to be the most symptomatic and obese. This suggests that the mild and moderate forms of the disease are underdiagnosed.¹⁸ Untreated sleep apnea is an enormous economic burden estimated to cost billions of dollars in additional medical costs in the United States.¹¹ Awareness and education needs to begin with medical students, residents, and primary care physicians, who are often the first to evaluate these patients. This training has been demonstrated to be effective. Haponik et al. showed that 82% of medical interns who were instructed in sleep medicine asked patients about their sleep patterns compared to 13% of medical interns who did not receive instruction.³⁴ Prioritizing screening for sleep apnea in the primary care setting has tremendous potential for identifying high-risk patients earlier, preventing or reducing adverse consequences, and relieving health care costs.

DISCLOSURE STATEMENT

This was not an industry supported study. Dr. Auckley has received research support from ResMed. The other authors have indicated no conflicts of interest.