To compare the short-term outcomes of acute knee injuries treated by specialists and generalists.
To compare the short-term outcomes of acute knee injuries treated by specialists and generalists.
Using patient logs, 168 adults with acute knee injuries were identified; 131 (78%) completed a questionnaire 3 months after initial presentation.
The mean age of the 77 male and 54 female responders was 34.6 years (range 18–73 years). The injuries were classified as mild (n = 35), moderate (n = 75), or severe (n = 21). Most responders were satisfied with their care and outcome, but 22% noted some functional limitations. The 59 patients seeing an orthopedist were more likely to have had a severe injury, more physician visits, activity limitations, lost time from work or recreation, and more pain when compared with the 72 patients who never saw an orthopedist. Excluding surgical patients, however, satisfaction was not significantly different by provider. After multivariate modeling (adjusting for age, sex, injury severity, and diagnosis), there was no significant association between having seen an orthopedist and either treatment success or satisfaction.
With the exception of time lost for recuperation in our community there is little difference in short-term outcome for patients with acute knee injury not undergoing surgery, regardless of the specialty of the treating physician.
As medicine becomes increasingly specialized, questions naturally arise as to the proper role of generalists and specialists in the management of specific problems. Many common injuries are cared for by both specialists and generalists. Provider decisions may be made by the patient, a primary care provider, or by a health plan. Little is known about the impact of the choice of initial provider on patient satisfaction and clinical outcome for common musculoskeletal injuries. These questions are now being asked at an accelerated pace. The clinical practice of medicine has changed considerably in the United States over the past 2 decades, with increasing emphasis on cost controls. These controls often take the form of restrictions on access to certain types of specialized care. Little is known about the impact of these restrictions on quality or patient satisfaction. Specifically, do generalists and specialists obtain similar clinical outcomes and levels of patient satisfaction for a given type and severity of injury (1–4)? Are referrals timely and appropriate?
We have taken advantage of a unique arrangement in our community in which all physician and hospital records are collated and available for review (5). This allowed us to compare satisfaction and outcomes of care provided by generalists and specialists for a common problem: acute knee injury. Outcomes were stratified by diagnosis and severity, and then compared between groups of patients whose care team did or did not include an orthopedic surgeon.
The medical care in our county is provided almost exclusively by 2 medical groups. Together these 2 groups operate all of the county's acute care general hospitals, emergency rooms, urgent care centers, and sports medicine clinics. Between March 1, 1997 and April 15, 1998, we concurrently identified all acute knee injuries presenting to the emergency rooms, urgent care centers, and sports medicine clinics at these facilities. Patient logs, maintained at each facility, were reviewed to ensure complete case ascertainment.
The complete medical records of each potential subject were also identified, through a registry that links the records of all medical care provided to residents of our county, regardless of the site or facility where care was provided (5). Records for each patient were reviewed to assure that the patient was qualified on the basis of the inclusion and exclusion criteria. Inclusion criteria were age > 17 years, a diagnosis of knee injury, resident of our county at the time of injury, injury within 60 days of presentation for care, and absence of other lower extremity injury. Exclusion criteria included an associated diagnosis of fracture; the presence of systemic illness likely to affect the knee, such as rheumatoid arthritis or osteoarthritis; and the presence of another concurrent injury to the lower extremities.
For each subject accepted to the cohort, all medical records were reviewed for relevant elements of past medical history and information regarding the diagnosis, severity, and treatment for the knee injury. All visits, whether to physicians or other medical providers, were reviewed extending from the index visit for a total of 3 months afterward. The study was approved by the institutional review boards of the 2 medical groups.
Diagnoses were sorted into 3 categories that appeared to cluster patients into similar treatment/prognosis groupings: Group I, sprain or strain, anticipated to be self-limited and responsive to symptomatic management; Group II, ligamentous or patellar soft tissue injury, likely to require more specific and intensive but still nonoperative treatment; and Group III, cruciate or meniscal injury, more likely to require surgery. These categories were selected after reference to the American Academy of Orthopaedic Surgeons Clinical Guideline on Knee Injuries (6), which uses these categories to advise primary care providers on diagnosis, treatment, and indications for referral. Each occurrence of a diagnosis was recorded, allowing the possibility to survey changes in presumptive diagnosis over time.
Injury severity was based on an algorithm of exclusion. Severe injuries were those meeting 1 or more of the following criteria at the time of the initial visit for the knee injury, as noted in the medical record: pain requiring narcotics, marked joint swelling, marked knee instability, or inability to bear weight on the affected leg. Moderate injuries were those in which pain was controlled by nonnarcotic medication, walking was possible but more strenuous activity was not, moderate swelling or instability were present, or the range of knee motion was limited. Any injury not qualifying as severe or moderate was considered mild.
Treatments recorded were grouped into modalities such as ice, rest, analgesics, physical therapy, joint injection or aspiration, arthroscopic surgery, and open surgical repair; they were then stratified by the category of injury event, final diagnosis, injury severity, and age of the patient.
All patients were contacted by letter 3 months after their initial visit and asked to complete the Medical Outcome Study Short Form-36 (SF-36) (7) and AAOS Sports/Knee (8) questionnaires. The SF-36 domains include pain, energy, mental health, physical function, physical role, emotional role, social function, health perception, and change in health. The AAOS Sports/Knee questionnaire domains include pain and physical function attributed to the knee, lost activity days, and satisfaction. Nonresponders were contacted by telephone. Responses were tabulated and scored as recommended for these instruments (8).
Demographic and diagnostic characteristics of responders were compared using a chi-square test for discrete factors and a Wilcoxon's rank sum test for continuous factors. Associations between patients having care from an orthopedist and discrete factors were evaluated using a chi-square of Fisher's exact test as necessary; univariate assessment with continuous factors, such as SF-36 subscales and patient satisfaction, were made with Wilcoxon's rank sum test or the exact version as appropriate. No multivariate models were constructed for the SF-36 subscales due to the extreme skew of their distributions. Spearman's rank correlation was calculated to evaluate the association between patient perception of treatment success and satisfaction. Multivariate modeling of success of care, as well as satisfaction with care, was performed using logistic regression. We adjusted for age, sex, and injury severity while assessing whether inclusion of an orthopedist in patient care was associated with success or satisfaction. To simplify the analysis, success was dichotomized as completely or very successful versus somewhat successful, not very successful, or unsuccessful; overall satisfaction was dichotomized as completely or very satisfied versus somewhat satisfied, dissatisfied, or very dissatisfied. A subset multivariate analysis including only patients not having surgery was also performed. Relative risk estimates (RR) and associated 95% confidence intervals (95% CI) were calculated, adjusting the odds ratio estimates from the logistic regression results using the method of Zhang and Yu (11). The alpha level was set at 0.05 for statistical significance.
A total of 3,160 medical records were identified and screened during the study period, of which 218 cases met our inclusion criteria. Fifty patients were subsequently excluded on the basis of 1 or more of our exclusion criteria. Thus, 168 patients were ultimately eligible for the study during the intake period, of whom 131 (78%) completed and returned the 3-month questionnaire. The demographic and diagnostic characteristics of nonresponders were not statistically different from the responders with regard to sex, injury severity, provider at first contact, or mechanism of injury. Nonresponders were more likely than responders to have had a diagnosis of sprain or strain (chi-square P = 0.005) and to be younger (nonresponders mean age 29.0 years, responders mean age 34.6 years, Wilcoxon's rank sum P = 0.0071).
Among the responders, there were 77 men and 54 women. The mean age was 34.6 years, with a range of 18–73. The mechanism of injury was a fall in 20 cases, a twist or bend in 39 cases, a sports injury in 52 cases, and unknown or unspecified trauma in 20 cases. The mean number of visits per patient was 6.0 ± 6.2, with a range of 1–34. There were, by our criteria, 36 mild, 74 moderate, and 21 severe injuries. Forty-one patients (31%) had only a single physician visit for diagnosis of their knee injury, and 20 patients of these 41 had only 1 visit of any type for their knee injury (15%).
Initial care was provided by family physicians in 22 cases, general internists in 8 cases, orthopedists in 18 cases, emergency room physicians in 52 cases, physiatrists (primarily in a sports medicine center) in 27 cases, and by other providers in 4 cases. The most common final treating physicians were physiatrists in 53 cases and orthopedists in 46 cases. Fifty-nine patients (45%) were seen by an orthopedist sometime during their care, whereas 72 (55%) patients were never seen by an orthopedist.
The frequency of final diagnosis groups by provider type is shown in Table 1. Inclusion of an orthopedist in a patient's care was significantly associated with diagnosis, with orthopedists more likely to have seen patients with meniscus or cruciate injuries and less likely to have seen patients in the other diagnostic groups (Wilcoxon's rank sum test P < 0.01). Inclusion of an orthopedist in a patient's care was significantly associated with injury severity: 31% of patients with mild injuries, 47% of patients with moderate injuries, and 62% of patients with severe injuries included an orthopedist in their care, (exact Wilcoxon's rank sum test P = 0.018). Patients seeing an orthopedist at some time were likely to be seen more frequently (median 7 visits [range 1–34] versus 2.5 visits [range 1–17]; Wilcoxon's rank sum test P < 0.0001); to have reported that their activities were more limited, with 0 being the least limitation and 100 being the maximum (median 18 [range 0–64] versus 0 [range 0–52]; Wilcoxon's rank sum test P < 0.0001); to miss more work days (median 1 [range 0–91] versus 0 [range 0–15]; Wilcoxon's rank sum test P = 0.0001); to miss more days of recreational activities (median 30 [range 0–90] versus 14 [range 0–90]; Wilcoxon's rank sum test P = 0.0018); to have had pain with activity, with 0 being least pain and 100 being maximum pain (median 32 [range 0–60] versus 16 [range 0–56] Wilcoxon's rank sum test P = 0.0033); and to have missed more days of work prior to returning to regular employment (median 2 [range 0–90] versus 0 [range 0–35] Wilcoxon's rank sum test P = 0.0169) If only those patients not having surgery are considered, only the difference in number of days of missed recreational activities is significant, with the median days missed being 25 for those seeing an orthopedist at any time during their care versus 14 days for those never seeing an orthopedist (P = 0.0105, for all other associations P > 0.05).
|Group I||37 (28)|
|Group II||27 (21)|
|Group III||30 (23)|
|Groups I, II||12 (9)|
|Groups I, III||11 (8)|
|Groups II, III||13 (10)|
|Groups I, II, III||1 (1)|
Univariate assessment of 5 SF-36 subscales (health perception, physical functioning, mental health, bodily pain, and energy) showed no statistically significant associations with the inclusion of an orthopedist in a patient's care (Wilcoxon rank sum test P > 0.05). Treatment for each patient was based on the recommendations of the physicians involved in that patient's care. The most common modalities were ice, rest, simple analgesics, and some sort of joint protection (i.e., compressive wrap, knee immobilizer, crutches, or cane.) Therapy, joint aspiration, and surgery were reserved for more specific and severe conditions. Surgery was advised in 33 cases and performed in 31 (all were arthroscopic procedures).
Patient outcome was classified in terms of pain, functional status, swelling, number of work days missed, days needed to return to normal recreational activities, knee function, and patient activity level. Association between provider pattern of care (never having seen an orthopedist versus having seen an orthopedist) and patient satisfaction with their care was significant (exact Wilcoxon rank sum test P = 0.039) (Tables 2 and 3).
|Completely satisfied||Very satisfied||Somewhat satisfied||Dissatisfied||Very dissatisfied|
|Nonorthopedist only||27 (34.6)||32 (41.0)||13 (17.0)||3 (3.9)||2 (2.6)|
|Referral to an orthopedist||7 (25.9)||4 (14.8)||15 (54.0)||2 (7.4)||0|
|Orthopedist only||7 (41.2)||6 (35.3)||3 (17.7)||1 (5.9)||0|
|Completely satisfied||Very satisfied||Somewhat satisfied||Dissatisfied||Very dissatisfied||Total|
|Orthopedist never seen||25 (37.3)||27 (40.3)||12 (17.9)||2 (3.0)||1 (1.5)||67|
|Orthopedist seen at some time||16 (28.1)||16 (28.1)||20 (35.1)||4 (7.0)||1 (1.8)||57|
At the time the questionnaire was completed, 102 (78%) patients reported no limitation in their knee function, 18 (14%) moderate limitation, and 10 (8%) severe limitation in their knee function. A total of 46 patients missed at least 1 day of work as a result of their knee injury. Of the 124 patients completing the question concerning overall satisfaction of treatment for knee injury, 41 (33%) answered that they were completely satisfied, 43 (35%) very satisfied, 32 (26%) somewhat satisfied, 6 (5%) dissatisfied, and 2 (2%) very dissatisfied. Assessing degree of treatment success among the 123 patients who completed the questionnaire regarding treatment success, 36 (29%) responded as 100% successful, 49 (40%) about 75% successful, 29 (24%) about 50% successful, 4 (3%) <50% successful, and 5 (4%) completely unsuccessful. The perception of satisfaction and treatment success were highly correlated, as expected (Spearman's rank correlation r = 0.77).
A multivariate logistic regression model was used to assess those patients who believed their care was <75% successful. Adjusting for age, sex, and injury severity, and given the numbers available, there was no significant difference in success between patients who did and did not have an orthopedist involved in their care (P = 0.1314; RR = 1.5; 95% CI 0.9–2.3). Adjusting for injury type yielded similar results (P = 0.2308; RR = 1.4; 95% CI 0.8–2.3). Assessing only patients not having surgery, the result was similar (P = 0.1683; RR = 1.6; 95% CI 0.8–2.5). Adjusting for injury type yielded similar results (P = 0.1625; RR = 1.6; 95% CI 0.8–2.5).
Patients who were somewhat or very dissatisfied with their care were compared with those who were completely/very satisfied using a multivariate logistic regression model. The model, adjusted for age, sex, and injury severity, found that patients who had an orthopedist involved in their care were significantly more likely to be less satisfied (P = 0.0171; RR = 2.0; 95% CI 1.1–2.8). After adjusting for injury type, the borderline significant result becomes nonsignificant (P = 0.1070; RR = 1.7; 95% CI 0.9–2.6). However, the estimate of the odds ratio is 1.7, indicating that patients who had an orthopedist involved in their care were still nearly twice as likely to be less satisfied. Assessing only patients not having surgery (n = 100), the differences in satisfaction were not significant (P = 0.19; RR = 1.6; 95% CI 0.8–2.7). Adjusting for injury type yielded nearly identical results (P = 0.1953; RR = 1.6; 95% CI 0.8–2.7).
In this study, we measured short-term clinical outcomes and patient satisfaction for acute soft tissue injuries to the knee joint in a community-based population. We have correlated those outcomes with patterns of care based on diagnostic grouping, severity, and the presence or absence of an orthopedic surgeon in the care-giving team. Knee injuries were chosen because in previous work we showed that knee injuries are cared for by a variety of providers, and that, although there is a trend for more severe injuries to be cared for by orthopedic surgeons and the more mild injuries to be cared for by nonorthopedists, this trend is far weaker for knee injuries than it is, for example, in shoulder or wrist injuries (9). Patterns of diagnoses, diagnostic changes, and treatment are similar in this study to those which had been observed retrospectively in our community (9, 10). Of note, most patients in our community are not bound by managed care restrictions and are free to chose providers. In addition, the density of orthopedic surgeons in our county is greater than 1/5,000, suggesting that access to orthopedic care is not limiting patient choice.
Our findings, that satisfaction and outcome were similar in the nonsurgical patients, whether or not an orthopedist was involved, suggest that provider specialty does not significantly contribute to patient satisfaction or functional outcome, at least in our patient population. Conversely, our results may reflect the ability of physicians and patients to choose the type of physician to optimize care based on the type of injury, the expertise of the physician, and the desires of the patient. The results were not affected by variation in severity or diagnosis. We are unable to determine whether or not the use of surgical intervention is optimal and whether any restrictions that limit access to surgical care might have adversely affected the outcomes and satisfaction in this group of patients.
The finding that patients having surgery have worse short-term outcomes and satisfaction should not be surprising, because most would still be recovering from surgery at the time of our followup survey, which asked about current symptoms rather than final expected status. We believe that in any study comparing outcomes between surgical and nonsurgical specialists, adjustments such as the timing of outcome assessments must be made for patients having surgery, so as not to skew outcomes such as cost or satisfaction toward nonsurgical care. Some adjustment may even be needed when comparing long-term outcomes, because surgical complications can again only occur in the surgical group. If one wished truly to compare overall outcomes of care of surgical and nonsurgical specialists, a prospective, randomized allocation of patients between the 2 groups would be needed. That was not done in the current study, but could be the subject of future investigation.
In conclusion, for patients without surgical intervention, we were unable to demonstrate that functional outcomes or satisfaction differed whether or not an orthopedic surgeon consulted in their care. Regardless of type of physician providing care, there is significant morbidity associated with acute knee injuries, even in a community-based population. Rather than concentrate on guidelines that require specialty referrals, further studies might profitably address methods to reduce the morbidity of these injuries, either by specific treatment algorithms or preventive education.