On the demand for telemedicine: Evidence from the COVID-19 pandemic.

Telemedicine can expand access to health care at relatively low cost. Historically, however, demand for telemedicine has remained low. Using administrative records and a difference-in-differences methodology, we estimate the change in demand for telemedicine experienced after the onset of the COVID-19 epidemic and the imposition of mobility restrictions. We find that the number of telemedicine calls made during the pandemic increased by 230 percent compared to the pre-pandemic period. The effects were mostly driven by older individuals with preexisting conditions who used the service for internal medicine consultations. The demand for telemedicine remained relatively high even after mobility restrictions were relaxed, which is consistent with telemedicine being an "experience good." These results are a proof of concept for policy makers to use such relatively low-cost medical consultations, made possible by new technologies, to provide needed expansion of access to health care.

INTRODUCTION

Telemedicine can be a powerful tool to expand the delivery of health care services at a relatively low cost (Bashshur, 1995; Ekeland et al., 2010). Even though technological innovations have eased the expansion of the supply of telemedicine services in recent years (Pandian, 2016), demand has remained stubbornly low (Wootton, 2008; Zanaboni & Wootton, 2012). The COVID‐19 pandemic seems to have changed that. The pandemic led many governments to impose lockdowns and social‐distancing policies, which induced patients around the world to experience telemedicine for the first time.

In this paper, we exploit the introduction and subsequent relaxation of social‐distancing policies to study how the demand for telemedicine changed in Argentina. We rely on administrative records from one of the largest providers of telemedicine in the country to build a panel data set that includes daily records of all calls received during 2019 and 2020. We use an event study and a difference‐in‐differences methodology to estimate the change in demand for telemedicine that happened after mobility was restricted at the onset of the COVID‐19 epidemic.

We find that the use of telemedicine soared. The total number of calls grew by 230%; the number of first‐time callers grew 198%. The number of resolved consultations increased by 235%. Calls referred to another specialist increased by 190%. Telemedicine calls resulting in prescriptions more than tripled, rising by 332%. These effects were driven mostly by older individuals with preexisting conditions who used telemedicine for internal medicine consultations. As spatial mobility began to increase, reaching levels that were 80% of those experienced prior to the pandemic, the use of telemedicine declined slightly but not quickly enough to fall to the level of use that was the rule in the pre‐pandemic period. We show that as mobility returned, a 1% increase in mobility resulted in a decrease in the use of telemedicine ranging from 0.8 to 2.5% points. Although more time is needed to fully assess whether demand will remain this high once mobility returns to pre‐pandemic levels, we take this result as preliminary evidence that the upward shift in demand is likely persistent.

Our paper exploits an external shock that triggered patients' learning about telemedicine. As is the case with many health care services, telemedicine can be characterized as an “experience good:” one that can only be accurately evaluated and compared to its substitute (in this case, in‐person visits) only after the product has been purchased and experienced (Andersen & Philipsen, 1998). By showing that demand dramatically increased after the onset of the COVID‐19 epidemic, and by presenting preliminary evidence of persistence of those effects, our results contribute to the large literature that analyzes patient learning about health‐care markets. For example, similar learning processes have be shown to reduce the costs of the drug‐patient matching process in pharmaceutical markets (Crawford & Shum, 2005), to increase the take up of new vaccines (Maurer & Harris, 2016), and to affect the choice of health‐insurance plans (Chernew et al., 2008).

This paper also contributes to the strand of the literature that studies how demand for new health‐care services is affected by exogenous changes in factors such as prices of services (Berman & Fenaughty, 2005), consumer's beliefs and social norms (Cranen et al., 2011), or the education and wealth of households (Chunara et al., 2020). We conjecture that the epidemic, by restricting access to traditional in‐person visits, induced consumers to overcome behavioral constraints to use telemedicine for the first time. Experiences with the service as a result of the pandemic seem to have led to a new equilibrium of higher demand, adding evidence to the literature on experience goods and the adoption of new technologies (Sunstein, 2019).

Finally, our results contribute to the new and expanding literature on the effects of the COVID‐19 crisis on the demand for services including online education (Ikeda & Yamaguchi, 2020), online retailers (Farrell et al., 2020), child care (Ali et al., 2020), and public transportation (Tirachini & Cats, 2020).

From a public‐health point of view, the importance of telemedicine as “forward triage” to sort patients before they arrive at the hospital has been of paramount importance during a pandemic. Telemedicine allows patients to be efficiently screened and directed to the most suitable health care provider, which effectively increases the capacity of the health‐care system (Hollander & Carr, 2020), and facilitates isolating those who may be infected by the virus. Our paper shows that telemedicine, when properly deployed and scaled up, can be relied upon as an important tool for public‐health management.

This paper is organized as follows: Section 2 reviews the literature on cost and benefits of telemedicine. Section 3 describes the setting in which the increase in the demand for telemedicine services took place. Section 4 shows how mobility declined in Argentina during the COVID‐19 crisis. Section 5 specifies the empirical strategy and the data used in the analysis. Section 6 presents the results while Section 7 discusses them. Section 8 concludes.

COSTS AND BENEFITS OF TELEMEDICINE

The terms telemedicine or telehealth are currently used to describe the provision of health‐care services remotely, by means of a variety of telecommunication tools including telephones, smartphones, and other devices, with or without a video connection (Dorsey & Topol, 2016). In recent years, the use of telemedicine has gained momentum, primarily because of the perceived potential to better distribute and control the use of medical services to improve the timeliness of delivery and, hence, the overall quality of health care. Indeed, telemedicine has been proven to increase the accessibility of health services, and to reduce travel time and related opportunity costs in the process of obtaining care (Bashshur, 1995). By providing access for war veterans and patients in rural areas, for example, telehealth offers an alternative to traditional health care by lowering the time and cost of receiving service (Jacobs et al., 2019; Sabesan et al., 2012). In addition, there is evidence that telemedicine is successful in reducing the need for ambulance transport, which could provide relief to the overcrowded health‐care system (Langabeer et al., 2016). Telemedicine can also increase the diversity of care to which an individual has access. As an example, for indigenous people living in remote areas, telemedicine provides an option that reduces the burden of travel and dislocation from community and family (Caffery et al., 2018).

The relative effectiveness of telemedicine, as compared to in‐person consultations, remains a subject of study. A scoping review of the use of telemedicine was inconclusive about the relative effectiveness of electronic and face‐to‐face consultations (Caffery et al., 2016; Roine et al., 2001). On the one hand, replacing traditional face‐to‐face patient care can potentially result in a breakdown of the traditional relationship between the health professional and the patient, leading to the potential for depersonalization of the service (Hjelm, 2005). On the other, patients who use telemedicine are overwhelmingly positive about the experience, reporting high quality and satisfaction levels (Jacobs et al., 2019; Kruse et al., 2017; Polinski et al., 2016). For many patients, electronic consultations were preferred for convenience and travel time (Donelan et al., 2019).

In spite of this, there is still a reluctance on the part of patients to increase their use of telemedicine (Wootton, 2008). There are several reasons why people may resist the use of new technologies such as telemedicine (Broens et al., 2007). First, there could be a general mistrust or lack of information among patients and health professionals about the effectiveness of telemedicine (Mair et al., 2007). Second, there are real – even if small – inconvenience factors, such as having to download and set up the technology, which could discourage its use, or trigger procrastination (Baicker et al., 2012; Bertrand et al., 2004; Kremer et al., 2019; Madrian, 2014; Rice, 2013). Third, a number of behavioral biases may limit the use of telemedicine. Individuals also may not download the applications required to use telemedicine services because of present bias, which makes people undervalue the future gains of having the service ready to use should they become sick (Kang & Ikeda, 2016; Kremer et al., 2019; Linnemayr & Stecher, 2015; Madrian, 2014; Williams et al., 2018). This can be compounded by optimism bias, which leads people to underestimate the probability of negative events, or by loss aversion, which can lead people to worry that using telemedicine could jeopardize access to in‐person visits later on (Kahneman et al., 1991). These biases build on a reticence by consumers to move from a known status quo to newer alternatives (Hartman et al., 1991; Kahneman et al., 1991; Rice, 2013; Suri et al., 2013; Tsai et al., 2019; Zhang et al., 2017).

It is also important to recognize other practical costs that telemedicine presents. With the adoption of telemedicine as a cheaper and more convenient alternative, there is the potential for excess health‐care utilization (Ashwood et al., 2017; Bavafa et al., 2018). That is, the question is still open as to whether any overall increase in demand from the use of telemedicine would reflect serving previously unmet health‐care needs, or overuse of health‐care services. Another concern is the possibility of overuse of prescriptions (Sprecher & Finkelstein, 2019). Similarly, legal and reimbursement issues could arise from limited or fragmented health‐care coverage through telemedicine services (Dorsey & Topol, 2016). The solution to these problems relies on the existence of a legal framework that appropriately regulates the use of telemedicine within the broader health‐care system.

In spite of these limitations, the health‐care community has encouraged the shift from an in‐person care model to a model of virtual care (Duffy & Lee, 2018). The COVID‐19 crisis highlighted the need for an easily deployable, mobility‐reducing, and low‐cost alternative to deliver care, especially to more at‐risk populations. The pandemic rapidly increased the perceived benefits of telemedicine. Moreover, mobility restrictions imposed by governments pushed individuals to use telemedicine, which in some circumstances may have been the lone, viable option for care. Mobility restrictions and concerns about the ramifications of exposure to COVID‐19 thus helped reduce the barriers associated with telemedicine, and resulted in more patients discovering that telemedicine could be a suitable substitute for in‐person care (Accenture, 2020).

TELEMEDICINE IN ARGENTINA

Before the onset of the COVID‐19 pandemic, the government of Argentina had already recognized the use of telemedicine as one of the main pillars of its strategy to ensure universal health‐care coverage. In 2019, the government launched its digital health strategy, which included among its goals the expansion of telemedicine as a tool to provide health services to geographically remote populations, improving accessibility, reducing the need for medical‐related transportation, and compensating for regional differences in access to health care (Gobierno de Argentina, 2019). When the COVID‐19 pandemic lockdown began, the government encouraged private health‐insurance providers to foster the use of telemedicine (Superintendencia de Servicios de Salud, 2020). At the onset of the pandemic, two providers of telemedicine services dominated the market in the country. One such provider was “Llamando al Doctor” (or “Calling the Doctor”) which offered services to health‐care providers, insurance companies, and individual patients across the country. At the end of 2019, the firm employed 108 doctors covering 11 medical specialties, including general medicine, pediatrics, and gynecology and obstetrics.

Patients access the service primarily through a mobile‐phone application, which asks a series of screening questions about: the medical specialty required, the reason for the consultation, and any existing health conditions the caller may have. Following the screening, the caller proceeds to the online consultation with a physician through a video call. Each video call can result in one of three outcomes: the first and most common outcome is that a doctor resolves the patient's issue during the online consultation. This was the case for 67% of the calls that were made in 2019. Doctors who resolved the issues presented in a call sometimes prescribed a medicine to the patient (as was the case in 11.5% of the overall calls). A second outcome is that a patient receives a recommendation to participate in a follow‐up call (as was the case for 8.3% of the overall calls). A third outcome is that a doctor refers the patient to an in‐person visit (as occurred in 10.8% of the calls).

Each call produces a log that registers the patient's gender and age, the medical specialty requested, a description of the reason for the call, and the diagnosis that resulted from the call. Table 1 provides some descriptive statistics for 2019 based on these (anonymized) administrative data. Patients that use telemedicine are relatively young (30 years old on average) and more likely to be women (57%). The majority of consultations relate to general medicine and pediatrics.

TABLE 1

Telemedicine service 2019: Descriptive statistics

		Proportion/Average	C.I/Std. Dev.	Obs.
Call outcome	Resolved	66.6%	[65.5%, 67.7%]	6890
	Prescription	11.5%	[9.7%, 13.3%]	1191
	Follow‐up	8.3%	[6.4%, 10.1%]	857
	Referral	10.8%	[9.0%, 12.7%]	1121
Medical specialty	General medicine	45.8%	[44.4%, 47.3%]	4740
	Ob/Gyn	18.5%	[16.7%, 20.2%]	1909
	Pediatrics	35.7%	[34.2%, 37.2%]	3691
Patient's characteristics	Age	30.6	15.1	10340
	Male	43.1%	[41.6%, 44.5%]	4455
	Previously diagnosed	25.0%	[23.3%, 26.7%]	2585

Note: In panel 1, the column “Proportion/Average” reports the proportion of calls that resolved the patient's issue, that ended in a prescription, that required a follow‐up call, or that referred the patient to another doctor. In panel 2, the column “Proportion/Average” reports the proportion of calls made to each medical specialty. In panel 3, the column “Proportion/Average” reports the average patient's age, and the proportion who were male, or who had a preexisting condition. The last two columns report the confidence interval (for proportion) or standard deviation (for averages) of each variable, and the number of calls/observations, respectively.

MOBILITY DURING THE COVID‐19 CRISIS

As a consequence of the COVID‐19 pandemic, governments around the world enacted different policies in an effort to contain the spread of the virus, and to minimize its socioeconomic impacts (Hale et al., 2021). These measures included mobility restrictions, economic relief programs, and investment in the health‐care system.

The pandemic started to unfold in Argentina at the end of February 2020. Media coverage and online searches of COVID‐19‐related terms started to increase the week of February 20 (the eighth week of the year). The first case of COVID‐19 was recorded on March 3.

Argentina's government issued the first social‐distancing measures on March 15 when schools were mandated to close. Three days later the government issued a formal, and strictly enforced, stay‐at‐home order for most of the country's population. By March 22, restrictions on public transportation were put in place. The timing and severity of the lockdown policies can be summarized using a Stringency Index. In Argentina, the index rapidly reached the maximum of one hundred by March 23, when the country reported the first four confirmed deaths by COVID‐19 (Hale et al., 2021). The government's fast and stringent response placed Argentina among the countries with the strictest lockdown measures in Latin America. De jure, these measures remained in place until November 2, 2020, when lockdown measures were formally relaxed.

These measures severely affected people's spatial mobility, especially at the beginning of the pandemic. We use publicly available data to build three indicators of observed mobility. The first two indicators come from the Apple Mobility Trends Report, which tracks driving and walking direction requests (Apple, 2020). The information shows the volume of directions requested on each date relative to the baseline volume in January 2020. A third indicator comes from Moovit, a company that provides a daily report of the use of its popular mobile application for public transit. This information shows the volume of requests received on each date relative to January 2020 (Moovit, 2020).

Figure 1 shows that the three indicators of mobility sharply declined around March 12, 2020. Even though many of the measures limiting mobility remained in place for months to come, walking and driving slowly and steadily increased over time. The use of public transportation remained depressed throughout the winter months (when lockdown measures were still in place), and started to increase after September. By the end of the year, mobility indicators were closer to their pre‐pandemic levels.

FIGURE 1

Mobility Indicators in Argentina for 2020. The figure plots indicators of spatial mobility. Walking and driving data were obtained through the Apple Mobility Trends Report, using baseline volume from January 13, 2020. The public transit indicator comes from Moovit, using baseline volume in the week of January 15, 2020. An indicator with value of 100 means that mobility on that day was the same as the reference date

EMPIRICAL STRATEGY

We use administrative records from “Llamando al Doctor” to construct a panel data set with the records of all calls received during the period from January 1 through December 31 for the years 2019 and 2020. These data allow us to analyse changes in the volume of daily calls received, and in the daily number of first‐time callers. We also observe the outcome of these telemedicine consultations: whether they were resolved, required a follow‐up call, or resulted in a referral to another specialist. In addition, we observe whether calls resulted in the issuance of prescriptions.

We estimate the effect of the onset of the COVID‐19 epidemic and the imposition of mobility restrictions on the demand for telemedicine using an event study and a difference‐in‐differences methodology. We define the onset of the pandemic (i.e., our ”treatment” date) as occurring during the 11th week of the year when mobility restrictions became into effect. Because mobility declined abruptly on March 12, we define a week to be a 7‐day period starting on each Thursday so that week 11 is the period from March 12 to March 18. A simple before‐and‐after comparison would not account for possible seasonal changes in demand for telemedicine. Thus, we compare outcome variables before and after mobility was restricted relative to their levels on the same dates in 2019.

We begin by estimating an event‐study model based on the following equation: where Y is the outcome variable measured at day d ∈ (1, …, 365) of week w ∈ (1, …, 49) of year t ∈ (2019, 2020). is an indicator variable equal to one if day d of year t belongs to week τ of that year, and Year2020 is an indicator for dates during the year 2020. The coefficients β estimate weekly changes in the outcomes for the period from January 1 to December 31 in 2020 relative to the outcome variable on the same dates of 2019. γ are week fixed effects, which control for seasonal trends, and δ are day‐of‐week fixed effects, which control for differences in the volume of calls received for different weekdays. The base week category is week 5, when the first news about COVID‐19 started to be discussed in the local media.

We also estimate average treatment effects according to the following difference‐in‐differences model: where Post is an indicator variable which equals one for dates during or after week 11 of year t. In this model, θ captures the average change in the outcome variable after mobility was restricted, relative to the same period in 2019. One key assumption of our difference‐in‐differences specification is that in the weeks prior to the onset of the COVID‐19 pandemic the outcome variable in both years followed the same trend. A test of this assumption is presented in Section 6.

RESULTS

We start by plotting the coefficients β and the 95% confidence intervals from the event‐study specification described in Equation (1). Figure 2 shows the results for two outcomes: the log of the number of daily calls (panel A) and log of first‐time callers (panel B). In both figures we overlay the times series of spatial mobility measured as the simple average of the three time series shown in Figure 1.

FIGURE 2

Event‐study Analysis. The green line shows the simple average of walking, driving, and public transit mobility indicators shown in Figure 1. Blue dots correspond to the point estimates obtained using Equation (1), and blue bars show the associated 95 percent confidence intervals. The vertical dashed line marks week 11, when mobility restrictions were first imposed. Weeks 6–8 are missing because telemedicine calls were not recorded those weeks in 2019. Week 17 is missing because telemedicine calls were not recorded that week in 2020

The event study allows to check for parallel trends during the pre‐treatment period. The coefficients for the number of daily calls and first‐time callers follow a flat pre‐trend before week 5, and start to increase somewhat 2 weeks before the mobility restrictions came into effect (when the risks of being infected were already being discussed in the media). For the six outcomes analysed in this study, we tested the following set of null hypotheses: and for j = {9, 10} (where j indexes the week). We only reject one null hypothesis (out of 12) at the 5% significance level. We take this as evidence that the parallel‐trend assumption is satisfied for our main outcomes of interest, even though some behaviors may have started to change a few weeks before the formal lockdown started.

By week 11, when mandated social distancing entered into effect and mobility dropped, the estimates for the daily number of calls and first‐time callers rose substantially. After that week, there was an upward trend in telemedicine use that reached a maximum during week 16. As mobility slowly started to converge to the pre‐lockdown values there was a mild decrease in the point estimates, which remained persistently higher than before the pandemic. These results point to an increasing demand for telemedicine that persisted even after the degree of mobility began to slowly return to pre‐pandemic levels.

Table 2 shows the results obtained with the difference‐in‐differences model described in Equation (2). In the months after the pandemic began, the number of calls grew by 230%, and the number of first‐time callers grew by 198%. The effect of mobility restrictions vary across call resolution outcomes. The largest effect was observed for calls that resulted in the issuing of prescriptions; these consultants grew by 332%. There was also a large increase in calls that required some type of follow‐up response (305%). Resolved consultations increased by 235%. The number of calls that resulted in a referral to another physician rose by 190%.

TABLE 2

Impact of mobility restrictions on telemedicine demand difference‐in‐differences estimates

	Main effects		Call resolution
	Calls	First‐time callers	Resolved	Prescription	Follow‐up	Referral
Post × Year2020	2.297***	1.976***	2.350***	3.324***	3.052***	1.899***
	(0.078)	(0.080)	(0.078)	(0.102)	(0.114)	(0.101)
Week F.E.	Yes	Yes	Yes	Yes	Yes	Yes
Day of week F.E.	Yes	Yes	Yes	Yes	Yes	Yes
Observations	690	690	690	690	690	690
Adjusted R 2	0.968	0.947	0.964	0.959	0.940	0.904
Average before Week 11 2020	30.58	18.42	19.28	5.77	2.52	3.56

Note: Each column presents the results of the difference‐in‐differences specification for a different dependent variable, estimating θ in Equation (2) using ordinary least squares. The dependent variables used in these models are (from left to right): log (number of calls), log (number of first‐time callers), log (number of resolved calls), log (number of calls resulting in prescription), log (number of follow‐up calls + 1), and log (number of referrals + 1). All models include week fixed effects (F.E.) and day‐of‐the‐week fixed effects (F.E.). The last line shows the average of the dependent variable in levels (i.e., not in logs) before the implementation of the mobility restrictions. * statistically significant at 10%, ** at 5%, *** and at 1%.

Heterogeneity. We next explore which types of patients were more likely to shift toward telemedicine. To that end, we estimated the coefficients in Equation (2) for different subgroups. Estimates of θ and 95% confidence intervals are presented in Figure 3. Point estimates are positive and statistically significant at standard confidence levels for all subgroups.

FIGURE 3

Treatment Effects: Heterogeneity. Panel (a) shows ordinary‐least‐squares estimates of θ in Equation (2) for the different subgroups specified on the x‐axis. “Baseline” reports the main results of Table 2. (For reference we place a horizontal dotted bar at that level.) “General Medicine”, “Ob/Gyn,” and “Pediatrics” show the effect on log (calls) related to general medicine, obstetric or gynecological care, and pediatric consultations. The categories labeled “< 18,” “18–24,”“25–39,”“40–54,” and “55–64” show the estimated effect on people in those age categories. “Male” and “Female” estimate the effect for patients of either sex. Finally “Preexist. Condit.” shows the estimated increase in calls by patients with a medical condition preexisting at the time of the call. Panel (b) depicts the same estimates when the outcome is log (number of first‐time callers). Blue bars report the 95 percent confidence intervals

The specialty with the greatest increase in demand was general or family medicine, which experienced an increase of 290% in the number of daily calls. The increase in the demand for telemedicine was greater among older patients than among younger patients; after social‐distancing measures were implemented, the number of daily calls made by those in the 55–65 age bracket increased by 367%, and the number of daily calls from those older than 65 grew by 406%. Calls from patients who reported having been previously diagnosed with a disease or illness increased by 296%. We find no differences between men and women in terms of the increased demand for telemedicine during the period.

Long‐run effects. If telemedicine is an experience good, then once people start using the service they might continue to use it in the future, even when the possibility of visiting the doctor in their office is again a possibility. Figure 2 shows that mobility steadily increased in the months after the initial lockdown period, and, contemporaneously, that weekly changes in the number of calls declined slightly throughout the same period. We exploit this variation to approximate an elasticity of demand of telemedicine to mobility by looking at the correlation between mobility and weekly changes in the outcome variables using the following regression: where the dependent variable is the weekly change in outcomes estimated for week τ using Equation (1), and ΔMob is the simple average of the three indicators of mobility during week τ of 2020 relative to January 2020 (as shown in Figure 1). We estimate this model for the full sample and for the weeks after week 11.

Table 3 shows the results. A 1% increase in the average mobility results in a reduction of between 0.8 and 2.6% points in the demand for telemedicine (measured by the number of calls). The estimates suggest that, despite traveling more, patients did not completely shift back to in‐person consultations; they were still opting for telemedicine as an alternative even at the end of the year when mobility was closer to pre‐pandemic levels. As mobility increased, the number of first‐time callers also decreased but only slightly. This suggests that both by adding new patients and by increasing the extensive margin, telemedicine may continue to expand, even after in‐person consultations are possible again.

TABLE 3

Demand for telemedicine and spatial mobility

	All 2020		2020 post week 11
	β Calls	β Callers	β Calls	β Callers
Δ Mob	−0.025***	−0.026***	−0.008*	−0.016***
	(0.004)	(0.003)	(0.004)	(0.003)
Obs.	48	48	42	42
Adjusted R 2	0.621	0.735	0.161	0.457

Note: Each column presents the ordinary‐least‐squares estimate of α 1 in Equation (3). Columns 1 and 2 are for the full sample. Columns 3 and 4 are estimated using only the weeks after week 11. * statistically significant at 10%, ** at 5%, *** and at 1%.

DISCUSSION

Using administrative records from one of the largest providers of telemedicine in Argentina and a difference‐in‐differences methodology, we find that the demand for telemedicine skyrocketed during the pandemic. The number of calls increased by 230%, and the number of first‐time callers increased by 198%. The number of first‐time callers as a share of the total dropped from 60% at the onset of the pandemic to 30% later on, a finding that is consistent with telemedicine being an experience good. The largest effect was observed on calls resulting in prescriptions; these calls increased by 332%. Calls that resulted in resolved consultations grew by 235%; calls that led to referrals to other specialists grew by 190%. The effects were driven mostly by older individuals with preexisting conditions who used telemedicine for internal medicine consultations.

The magnitude of our results is in line with similar findings in the literature from elsewhere and under different circumstances. Barnett et al. (2018) find that the use of telemedicine increased 261% as a result of expanded coverage of telemedicine services by Medicaid and private insurers in the United States. Harvey et al. (2019) find a 30% increase in the use of telemedicine as a result of expanded insurance coverage for telemedicine through state‐level legislative changes. Park et al. (2018) report a 228% increase in telehealth video calls when video consultation was integrated as a service. Der‐Martirosian et al. (2020) find an increase in the use of telemedicine of 50–205% in the context of a natural emergency (the 2017 Atlantic Hurricane season).

Our paper is the first one to document the existence of an untapped demand for telemedicine services in a middle‐income, developing country. We also show that the increase in the use of telemedicine slightly decreased after mobility restrictions eased, but not enough to undo the increase in demand. This result is consistent with survey evidence of more than 2700 patients (conducted in China, France, Germany, Japan, the United Kingdom, and the United States) which showed that about 60% of respondents said that, based on their experiences with telemedicine during the pandemic, they want to keep using the technology (Accenture, 2020). We take our results as tentative evidence suggesting that the demand for telemedicine did not decline sharply when mobility greatly increased. However, our study is somewhat limited to directly inform about the post‐pandemic long‐run trends in the use of telemedicine. More time and data are needed to better assess how permanent the increase in demand is.

In principle the increase in the use of telemedicine in our specific setting could be explained by two channels. First, an increase in the overall number of consultations (both in person and remotely) could to some extent stem from the pandemic itself. The number of COVID‐19 cases, however, remained fairly low for 3 months after the lockdown started. Second, the increase could result from a substitution effect by which patients switched from in‐person to telemedicine consultations.

How much can telemedicine substitute for in‐person appointments? Unfortunately, our data do not allow us to observe the number of in‐person consultations held; such observations would be needed to disentangle these two channels. We offer, instead, tentative evidence consistent with telemedicine partly substituting in‐person consultations.

First, the share of first‐time callers in the total number of telemedicine calls dropped from 60% at the onset of the pandemic to about 30% by the end of 2020. This is consistent with the conjecture that telemedicine is an experience good. Patients might have been hesitant to use telemedicine for the first time but, as they started using the service, they continued to do so throughout the year. In fact, about 65% of patients in our data are recurrent users of telemedicine, meaning that they called multiple times after their initial call.

Second, in order to be able to more directly assess the substitution between telemedicine and in‐person appointments, we obtained data on the number of these two types of appointments from a health‐insurance provider. We find that, initially, as mobility restrictions were imposed, patients seemed to make fewer in‐person appointments while at the same time they increased their telemedicine consultations. As mobility increased the number of in‐person consultations increased and telemedicine decreased somewhat. At least so far, however, the degree of telemedicine appointments seems to have stabilized at a higher level than the one observed before mobility was restricted.

CONCLUSION

Prior to the COVID‐19 pandemic, many governments and the World Health Organization viewed telemedicine as a tool that could be used to increase access to health care, reduce health‐care costs, and expand service, particularly to geographically remote and underserved populations (WHO, 2016). The COVID‐19 pandemic made even more clear the need to adopt innovative solutions that can provide relief to strained health‐care systems, help meet increasing demand, and minimize the risk of transmission of disease. This paper is a proof of concept that there was a hidden demand for telemedicine, and that policy makers have space to foster and accelerate the adoption of technological solutions to make health care delivery available to more people (Tanriverdi & Iacono, 1999). Behavioral tools could help lower barriers to the service, and nudge people into using it, and lead a wider array of people to reap its benefits. Providing patients the ability to experience the service could go a long way toward ensuring wider, sustained use to meet growing needs.

CONFLICT OF INTEREST

No conflicts of interest are declared by the authors.

TABLE A1

Test of Parallel Trends

	Main effects		Call resolution
	Calls	First‐time callers	Resolved	Prescription	Follow‐up	Referral
H 0 : β t1 = β t2	0.676	0.321	0.663	0.122	0.242	0.950
β t3 = β t4 = 0
H 0 : β t9 = β t10  = 0	0.022	0.131	0.141	0.441	0.216	0.840

Note: The table shows the p‐values of Wald tests of the joint hypotheses specified in the first column, where β indicates the j − th week (with j = 1, ..4, 9, 10).

TABLE A2

Robustness check for Difference‐in‐differences Estimates: Excluding a 3 week window pre and post week 11

	Main effects		Call resolution
	Calls	First‐time callers	Resolved	Prescription	Follow‐ up	Referral
Post × Year2020	2.465***	2.106***	2.504***	3.337***	3.217***	1.930***
	(0.074)	(0.080)	(0.078)	(0.122)	(0.128)	(0.117)
Week F.E	Yes	Yes	Yes	Yes	Yes	Yes
Day of week F.E	Yes	Yes	Yes	Yes	Yes	Yes
Observations	616	616	616	616	616	616
Adjusted R 2	0.971	0.952	0.967	0.958	0.944	0.912
Average before week 11 2020	30.58	18.42	19.28	5.77	2.52	3.56

Note: As a robustness test by excluding 3 weeks before and after week 11. Each column presents the results of the difference‐in‐differences specification for a different dependent variable, estimating θ in Equation (2) using ordinary least squares. The dependent variables used in these models are (from left to right): log (number of calls), log (number of first‐time callers), log (number of resolved calls), log (number of calls resulting in prescription), log (number of follow‐up calls + 1), and log (number of referrals + 1). All models include week fixed effects (F.E.) and day‐of‐the‐week fixed effects (F.E.). The last line shows the average of the dependent variable in levels (i.e., not in logs) before the implementation of the mobility restrictions. * statistically significant at 10%, ** at 5%, *** and at 1%.

TABLE A3

Demand for Telemedicine and Spatial Mobility: Other Outcomes

	β Resolved	β Prescription	β Follow−Up	β Referral
Panel A: All 2020	−0.028***	−0.043***	−0.031***	−0.015***
	(0.003)	(0.003)	(0.005)	(0.004)
Panel B: 2020 post week 11	−0.013***	−0.023***	−0.003	0.010**
	(0.002)	(0.004)	(0.005)	(0.005)

Note: Each column presents the ordinary‐least‐squares estimate of α 1 in Equation (3) for a different outcome. The top panel present results for the full sample. The bottom panel presents results estimated using only the weeks after week 11. * statistically significant at 10%, ** at 5%, *** and at 1%.

TABLE A4

Average Call's and Patient's Characteristics

	Llamando al	Health insurance
	Doctor sample	Provider sample
Resolved	62.2%	59.9%
Prescription	58.9%	69.5%
Follow‐up	25.3%	32.9%
Referral	8.1%	1.7%
General medicine	74.5%	82.2%
Ob/Gyn	14.0%	13.3%
Pediatrics	11.5%	4.5%
Age	40.4	48.2
Male	39.5%	48.2%
Pre‐existing condition	46.1%	56.9%

Note: The table compares the average characteristics of the outcomes of the calls, the types of calls, and the characteristics of callers of the “Llamando al Doctor” sample (column 1) with the subsample of patients enrolled with one health insurance provider from which we secured data (column 2).