Who Uses Telehealth? Setting a Usage Baseline for the Early Identification of Pandemic Influenza Activity

Introduction

In response to the 2009 A/H1N1 pandemic, it was suggested that nonroutine surveillance strategies could enhance existing influenza surveillance infrastructures. ¹ The most commonly known source of nonroutine surveillance data is the Google Flu project, which combines influenza-like illness physician reporting with influenza-like illness–related online search queries as an early warning system for influenza activity. ^{2 –4} Some jurisdictions such as the United Kingdom, ⁵ France, ⁶ Ireland, ⁷ and Australia ⁸ have used health visit and/or teletriage data to supplement their routine influenza surveillance portfolio during the pandemic, with conservative success.

Similarly to these aforementioned jurisdictions, Ontario, Canada has its telehealth program, which has been investigated as a potential source of influenza surveillance enhancement. ^{9 –11} The evidence collected from this nonpandemic research suggests that Ontario Telehealth could be used in a similar fashion to augment existing influenza surveillance, in both pandemic and interpandemic years. However, in order to use any data for enhanced surveillance during a pandemic, it is critical to understand how the data behave under normal (i.e., interpandemic) circumstances—both in terms of volume patterns as well as in terms of user patterns. Only once “normal” patterns are established can aberrant patterns be more confidently identified. This concept has been suggested by Olson et al., ¹² but little work has been done surrounding its application. In an effort to provide a telehealth baseline for respiratory conditions, this article describes respiratory call patterns made to Ontario Telehealth during the post–severe acute respiratory syndrome prepandemic period spanning from 2004 to 2006.

Methods

Twenty-five months of Ontario Telehealth data spanning June 1, 2004–June 30, 2006 were secured as part of a larger project to assess the suitability of Ontario Telehealth for influenza surveillance in Ontario. ¹³

Telehealth classifies all calls into one of three categories: (1) Health information; (2) service referral; and (3) symptom.

Health information calls are calls where someone requests general health information (in the absence of symptoms). Service referral refers to calls made to telehealth for referral to specialist services; this is a service made available to people living in underserviced areas. Finally, calls where health advice is provided through the application of a clinical decision tree are classified as symptom calls. For these calls, the recorded age is that of the patient, not that of the caller.

From these data, only symptom calls that were respiratory in nature were reviewed. In order to be classified as an acute respiratory symptom call, an Ontario Telehealth call's final algorithm was required to be one of the 27 listed in Table 1. These algorithms were preselected by an emergency medicine physician with extensive knowledge of syndromic surveillance systems and were selected specifically to meet this overarching study's aims.

Descriptive analyses were conducted on the respiratory subset. These included a general description of the consultation patterns, followed by cross-tabulations describing the seasonality and time of day patterns, followed by the generation of annual and average weekly consultation rates. Using the Canadian 2006 Census as the standard population, age-standardized per 1,000 population consultation rates were also calculated. Finally, service use patterns by provincial quintile were also determined. Using the 2006 Census average household income variable, five quintiles were calculated for the complete province of Ontario, based on a method devised by Roos and Mustard. ¹⁴ The lowest quintile (1) represented the highest average income range, and conversely the highest quintile (5) represented the lowest income range. Each Ontario Telehealth respiratory call for which a forward sortation area (FSA) (first three characters of a six-character postcode) was available was then assigned its corresponding quintile. The income breakdown by quintile was as follows: Q1, ≥CAN$96,510; Q2, CAN$78,675–CAN $96,509; Q3, CAN $68,477–CAN $78,674; Q4, CAN $60,672–CAN $68,476; and Q5, <CAN $60,672.

Statistical significance was ascertained by conducting chi-squared analyses (for descriptive statistics).

Results

During the 25 months under study, in total, 2,042,450 calls of all types were made to Ontario Telehealth. The largest proportion (approximately1.7 million) was symptom calls (83.8%). Of these symptom calls, 291,066 (17.0%) were extracted to create a “respiratory” subset (daily mean, 383; 42.8% male, 54.7% female).

In general, respiratory symptom calls showed a seasonality pattern in calls consistent with published influenza activity patterns for Ontario ^{15 –18} : An increase in calls that peaked annually in January/February (monthly mean, 11,643 calls; range, 7,088–13,088) for the two seasons under study (Fig. 1). In terms of weekly patterns, on average, approximately 41,600 respiratory calls were made per day of the week to Ontario Telehealth, with exceedances during the weekend, as well as a slightly above-average call rate on Mondays (p<0.0001).

OPEN IN VIEWER

Fig. 1.

Frequency of Ontario Telehealth respiratory calls by date, June 1, 2004–June 30, 2006.

Looking at the age-standardized per 1,000 population, Ontario Telehealth respiratory symptom consultation rates suggested an inverse relationship between the age group of the caller and the consultation rate, with a small exception in the oldest age group (>65 years), which showed a small increase relative to the 45–64-year age group (Fig. 2). The Ontario Telehealth respiratory rate of calls made for all children under the age of 5 years was 158.4/1,000 population. On average, the mean weekly all-ages consultation rate was 0.21/1,000 population, with a range of between 0.11 and 0.43/1,000 population respiratory calls per week. As a general rule, Ontario Telehealth peaks in call volumes were approximately twice the mean number of calls made to Ontario Telehealth for respiratory complaints during the period under study.

OPEN IN VIEWER

Fig. 2.

Annual Ontario Teleheatlh respiratory call rate per 1,000 population by sex and age group.

For the sex-stratified patterns, the male rates exceeded the female rates for the younger age groups, whereas females exceeded males for the older age groups, although the magnitude of calls in older age groups was inferior to the magnitude of calls for younger age groups, especially those under 5 years of age.

In total, 276,288 respiratory calls made to Ontario Telehealth from June 1, 2004 to June 30, 2006 were assigned a valid Ontario FSA (95.0% of all respiratory calls). The remaining (nonassigned) FSAs either were for out-of-province callers (e.g., visitors) or blank or were in a format inconsistent with FSA nomenclature. Consultation rates per 1,000 population per year stratified by age group and by quintile and consultation rate ratios for the Ontario Telehealth datasets are provided in Table 2.

For individuals 45 years of age and above, being in a higher quintile (i.e., lower income bracket) meant that the individual was more likely to go to call Ontario Telehealth. Conversely, individuals from lower quintiles (i.e., higher income bracket) were less likely to call Ontario Telehealth. However, those 5–14 and 15–44 years of age showed a bimodal relationship between income and consultation rate, with Q1 and Q5 having the highest consultation rates.

The highest consultation rates were with children aged below 5 years of age, whereas the widest consultation rate ratio range was for individuals 45–64 years old.

Discussion

From these data, it is clear that the advent of annual respiratory illness seasons resulted in surge capacity. During peak activity, weekly Ontario Telehealth call rates were up to more than twice the weekly mean and up to four times as high as the lowest weekly rate. Data such as these can and should be used for exercises such as seasonal and pandemic forecasting.

The overrepresentation of younger age groups relative to older age groups may be partially explained by three factors: (1) Younger children are more prone to respiratory ailments than their older counterparts. (2) Calls for younger children are made by parents who are more likely to use telehealth than their older counterparts. ¹⁹ (3) Younger parents are more likely to seek out care for their children than older adults with more parenting experience. In terms of the more pronounced overrepresentation of females in the older age groups, this is likely due to the established pattern that women tend to use medical services more than their male counterparts. ^20,21

The overall relationship between income and consultation rate was linear, with the exception of the 45–64- and >65-year age groups, suggesting that there may have been factors other than income and age that contributed to the consultation usage for Ontario Telehealth. As to why this was the case, it is difficult to ascertain without conducting a regression analysis to control for factors other than income. However, it is somewhat surprising given that the Canadian literature pertaining to the use of healthcare services suggests that individuals in lower-income neighborhoods are more likely to seek ambulatory care ²² or require hospitalization. ²³ Nonetheless, this pattern was consistent with the patterns identified by Hogenbirk et al. ¹⁹ in their evaluation of the Ontario Telehealth pilot project in 2002. Furthermore, the use of the FSA rather than a full postal code or a dissemination area may result in the homogenization of income over what are sometimes large, economically heterogeneous spatial areas. The availability of full postal codes would allow for a broader determinant approach using a deprivation index such as the one developed by Pampalon et al., ²⁴ which, in turn, could allow for the better analysis and understanding of utilization patterns.

While the syndromic surveillance system has primarily interested itself in establishing a baseline and monitoring exceedances to this baseline, the recent pandemic experience has showed us that not only are total increases in patient volumes important and indicative of unusual activity, but the demographic makeup of the patient population can also change and be indicative of activity of concern. During both the recent A/H1N1 pandemic as well as during the Spanish influenza pandemic, ^{25 –28} the obvious usual suspects (young and elderly) did not get sick in the same proportions relative to the proportion of otherwise healthy adults who got sick. Teletriage data can and should be used not only to monitor overall exceedances in usage, but also to monitor deviances from established demographic patterns. This would provide another level of monitoring that the limited number of cases reported through routine surveillance channels currently precludes doing. Furthermore, the added cost of this approach would be minimal because of the availability of near-real time electronic data collected by Ontario Telehealth.

It is important to remember that data from a Phase I (retrospective) evaluation, such as the one described here, need to be routinely reevaluated/updated in a Phase II (prospective) application so as to ensure that the Phase I findings continue to be relevant. With the recent upswing of Internet-driven health data, it is likely that teletriage usage patterns have changed as a result of these alternate avenues for advice. Consequently, the Ontario Telehealth system described here, if prospectively monitored, should only use the results reported here as a starting point from which to prospectively monitor respiratory activity.