Virtual Visits for Upper Respiratory Tract Infections in Adults Associated with Positive Outcome in a Cox Model

Introduction

Upper respiratory tract infections are common viral illnesses, which make up a large proportion of visits to urgent care centers. In 2012, of the 160 million total number of urgent care visits nationally, 29% were for upper respiratory tract infections. ¹ With the growth of telemedicine, some of this burden has shifted, but we do not have much evidence about the quality of a virtual visit compared to one in urgent care. In addition, identification of other factors that adversely affect outcome of upper respiratory tract infections may help in defining prognosis and therapeutic strategies.

Telemedicine care for low acuity urgent care conditions has been in existence for years, but there have been few studies looking at its prognostic ability or effect on the quality of care. Findings were generally favorable with respect to quality. Of 526 woman diagnosed and treated with antibiotics for urinary tract infections on telemedicine, 95% were followed up and showed results similar to what would be expected for in-person care. ² Two studies found that telemedicine diagnosis and evaluation were comparable to office acute pediatric care, but outcomes were not evaluated. ^3,4 A 2010 Cochrane meta-analysis included several trials involving more than 800 people. According to the authors, these studies appeared to be well conducted, but patient numbers were small in all but one. Although none of the studies showed any detrimental effects from the interventions, neither did they show unequivocal benefits and the findings did not constitute evidence of the safety of telemedicine. None of the studies included formal economic or statistical analysis. All the technological aspects of the interventions appeared to have been reliable and to have been well accepted by patients. However, one trial was concerned with telemedicine in the emergency department, one with video consultations between primary healthcare and the hospital outpatient department, and the remainder dealt with the provision of home care or patient self-monitoring of chronic disease. ⁵

Unlike some who have reached favorable conclusions regarding the literature and telemedicine outcomes for urgent care, ⁶ our findings supported Whitten et al. ⁷ regarding the inconsistency in documenting the effectiveness of telemedicine across a wide range of disciplines. Studies on the most common, low acuity, acute clinical problem seen virtually, namely viral upper respiratory tract infections, have not been done in the adult population. We wanted to begin addressing this through a retrospective Cox model study, which is a sound and practical design method for analyzing viral upper respiratory tract infections and following them forward to an outcome. ⁸

Materials and Methods

Southwest Medical is a multispecialty group in Las Vegas, Nevada with more than 330 providers, including physicians and mid-level practitioners. It is a part of OptumCare and is the primary medical group for more than 330,000 commercial health plan members. We examined a cohort of 6,756 patients with newly diagnosed upper respiratory tract infection. We used health plan claim data to identify all patients seen in urgent care and telemedicine for new viral upper respiratory tract infections (based on a collection of International Classification of Diseases [ICD-9] codes corresponding to the following: upper respiratory tract infection, common cold, sinusitis, bronchitis, pharyngitis, cough, and nasal congestion) between January 2014 and September 2014 and collated their comorbidities. These 12,566 cases were tracked for follow-up in any clinical setting—urgent care, telemedicine, office, emergency department, hospital—within a 2-week period of initial diagnosis. The accuracy of our data acquisition was scrutinized by reviewing the electronic medical records (TouchWorks by Allscripts) of these 12,566 to confirm the initial diagnosis of upper respiratory tract infection and verify that follow-up within 2 weeks was specifically related to this initial diagnosis. To establish a cohort of incident cases, we included only those with commercial insurance as telemedicine was exclusively available to these member holders during the study period, and only those 18–64 years of age. We excluded the older since most seniors had Medicare products, which were not yet eligible for the benefit of virtual visits. We did not include the pediatric age group as we wanted to focus solely on adults in this study.

Defining the quality of care is a difficult but essential early step. Mainz wrote that it was “the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current knowledge.” ⁹ Quality indicators in primary care have been error-prone, but using consensus techniques, which systematically combine evidence and opinion with guideline-driven approaches, facilitates quality improvement. ¹⁰ We could not find an existing definition that was both appropriate and validated, so we created a functional indicator for upper respiratory tract infections as follows: clinical quality for an upper respiratory tract infection is inversely proportional to the risk of follow-up, if any, within 2 weeks of the initial visit. Repeat visits have been universally considered a measure of low quality and an added cost of care. ¹¹

We chose to look first at claims data for the sole reason that it allowed the easiest way to reliably track follow-up within 2 weeks of a visit for an upper respiratory tract infection. However, we recognized that this source of metrics is grossly inadequate for incentives, transparency of cost and effectiveness, and continuous quality improvement. It should be augmented with detailed clinical outcome data that (a) offer more valid measures of performance and (b) enable researchers to establish and develop evidence-based practice guidelines. ¹² Random sampling of medical records found cases that had no documented follow-up at all and some that had no relationship to the initial visit for an upper respiratory tract infection. So we chose to review all 12,566 claim data cases of upper respiratory tract infection and selected 6,756 which were deemed true failures since they received follow-up within 2 weeks due to worsening condition, development of a complication, or were initially misdiagnosed.

Our main independent variable selected was a Virtual Visit. The telemedicine clinic for virtual visits was established in 2014 on a 24/7 basis for low acuity acute clinical conditions as a benefit for health plan members. It was staffed by salaried providers, both physicians and extenders. They used company network computers/laptops. Audio and visual were both in high definition. Some were based in Las Vegas, while others worked from their homes across the country. During peak hours, up to four providers were available to take calls for these virtual visits. We used the American Well platform to conduct these virtual visits. At start-up, high-quality audiovisual visits occurred 75% of the time, and after a few months, we achieved a steady rate of 95%. Providers had direct access to live patient electronic medical records during the virtual visits.

Patients independently chose to seek virtual care through a computer log-on or mobile phone application. This did require prior sign up through the vendor, American Well. During week day hours, a medical assistant acted as a virtual clerk who checked in patients. At other times, patients connected directly to the provider that they selected by clicking on the appropriate icon.

The in-person urgent care visits occurred in any of the existing 12 brick and mortar clinics in the Las Vegas valley. These ranged from one 24-hr facility, which included the ability to do cardiac and other acute medical workups, and testing with comprehensive laboratory menus, X-ray, computed tomography, and ultrasound. There were also smaller clinics meant for lower acuity care, only week day hours, and minimal point-of-care testing capabilities. Collectively, these urgent care clinics were staffed with various combinations of physicians and extenders on shifts from a solo provider to up to five overlapping depending on the site.

During the study period of 9 months, differences existed between virtual and in-person visits. Connection time to a virtual provider averaged 4 min with visit length averaging 8 min. The in-person visits required an average of 40 min before the patient was seen by the urgent care provider. The face-to-face time with the providers lasted <2 min. Co-pays were 40% less for virtual visits.

Explanatory variables or modifiers were selected since they were considered to have potential negative effects on upper respiratory tract infection outcomes: age, antibiotic given for the upper respiratory tract infection, and presence of chronic disease (chronic obstructive pulmonary disease, asthma, coronary artery disease, active cancer, immunologic/rheumatologic disease), and immunosuppressive medication. We determined the prevalence of virtual visits for the initial upper respiratory tract diagnosis and examined for differences between patients who did and did not have a virtual visit at baseline.

Survival analysis is an approach that uses a collection of statistical procedures for data analysis, in which the outcome variable is time until an event occurs. ^13,14 This is a validated way to look at times to an event such as follow-up within 2 weeks of an initial visit for an upper respiratory tract infection. Cox proportional hazard model was selected as the specific statistical method as it has been widely used and is semiparametric as the baseline hazard function does not have to be specified. ¹⁵ The Cox model assumption that the hazard functions for two individuals at any point in time are proportional and that there is no multicollinearity among covariates were tested by plotting the Kaplan–Meier Survival curves together. We also did the more complex complimentary log–log plot. A plot of the logarithm of the negative logarithm of the estimated survival function against the logarithm of the survival time ought to produce parallel curves if the hazards are proportional across the groups. ¹⁶ We used XLSTAT-Medical, Addinsoft, 2015 due to low cost, complete statistical bundle, ease of use, and its seamless integration with Excel, which held our raw data.

In a sensitivity analysis, we examined data in the first quartile, then the third, to see if there was a difference that might be attributed to start-up factors. For example, there might have been a predilection toward lower acuity cases early on. In contrast, providers' limited telemedicine experience might have led to more in-person follow-up.

Results

The median age of our 6,756 visits for viral upper respiratory tract infection was 41 with an average age of 40.3. Ninety-six percent were seen in person and had the same median and average ages. The remaining 6% who were seen virtually were younger with median age of 37 and an average of 38.3 (Table 1). On multiple regression analysis, virtual visits were more common in younger adults with an odds ratio of 1.02 (p = 0.02). Although the majority of the virtual cases were female (61%), this was less than the 65% for the in-person urgent care group.

Other demographic data were ambiguous. Of the virtual cases, 37% seen were Caucasian, whereas there were 46% in the urgent care group. However, many patients declined to declare a race, and many others had no clear reason to go without a designation. Information on religion, educational level, and income could not be obtained by health informatics despite many attempts.

Of the 383 virtual cases, only 4% had follow-up within 14 days of initial diagnosis. Of the 6,373 in-person urgent care ones, 26% had follow-up (Z score 9.95, p < 0.0002). For those initially seen virtually, the average days to follow-up of 2.0 (maximum of 14) was greater than the 1.6 days for in-person urgent care visits. The plot of Kaplan–Meier curves showed no crossing over, which indicated that the hazards were proportional across the groups (Fig. 1). Survival (or not being followed-up within 2 weeks) of virtual visit patients was unchanged after adjustment for other prognostic factors of age, antibiotic treatment, immunosuppressive medications, chronic medical conditions, and other descriptive factors. Cox proportional hazard ratio for follow-up was 0.552 (confidence interval of 0.324–0.939, p = 0.028) in these virtual visit cases (Table 2). This means that a patient diagnosed virtually with upper respiratory tract infection had a 45% reduced chance of receiving follow-up within 2 weeks compared to those diagnosed in person at an urgent care. There was no difference when the model was tested with combinations of other explanatory variables. None of these variables achieved statistical significance when run independently in the Cox model. Finally, these findings were unchanged when the data were run for first quarter and the third quarter time periods.

OPEN IN VIEWER

Fig. 1.

Kaplan–Meier curves for virtual (telemedicine) and in-person (urgent care) visits.

Discussion

Our study demonstrated that among a cohort of upper respiratory tract infection patients, a virtual visit can be considered an independent prognostic factor. This is a unique finding as we have not discovered any previous literature that has addressed this. However, we can still consider some possible explanations. The study finding that there was significantly less follow-up for upper respiratory infections seen on a virtual visit suggests more accurate initial diagnosis. In addition, the longer time spent with the patient virtually likely allowed for more education about the condition and indications for appropriate follow-up. Reassurance that the condition is self-limiting can help reduce unnecessary follow-up visits of any type. Antibiotic prescription rate of 25% for viral upper respiratory tract infection cases seen virtually was comparable to the very favorable one of 21% for in-person urgent care cases, so that this could not substantially account for the higher average patient satisfaction score of 95% for virtual visits compared to 84% in urgent care.

As this is an observational study, there are limitations. There are well-known problems with the validity and reliability of claims data. However, we justified starting with this data set as it was a practical way to capture the follow-up in different clinical settings of those diagnosed with our index disease. By doing an extensive electronic medical record review, we were able to select appropriate cases for our Cox analysis. Although we adjusted for possible prognostic factors, there is no literature evidence that these are known variables that affect outcomes of patients with upper respiratory tract infections. Even though the Cox proportional hazard model is not perfect at mimicking reality, we did meet its assumptions. In addition, there may have been bias of lower acuity cases seeking care virtually although the two groups were demographically similar. Also, we were not able to get some common demographic data that may have impacted our results. However, from the outset, our study was going to be unmatched, using the Cox model to examine the relationship between the survival (i.e., not being followed up for an initial upper respiratory tract infection) of a patient and several explanatory variables. Provider demographics were not considered in detail. Experience, age, sex, and provider type are some traits that may have an impact on follow-up rates.

The increase in time spent by the provider with a patient on a virtual visit, when compared to in-person, may account for a large part of our findings. This should be controlled for in further research. It may be that with greater utilization of the virtual clinic, this difference will diminish spontaneously.

We believe that efforts to expand utilization of virtual care for adult patients with upper respiratory tract infections should be considered. Studies looking at other time periods can help validate our findings. Other age groups and diagnoses would also be worth looking at to see if virtual visits have a similar positive prognostic impact. Along with more convenience and lower cost to the patient, this evidence supporting higher quality of care for viral upper respiratory tract infections in adults should be an added impetus for action.