Abstract
Objectives:
To identify rates of telemedicine provision during the COVID-19 pandemic and predictive institutional factors among 4-year and graduate colleges and universities.
Participants:
The study (n=364) included the websites (.edu) of accredited public non-profit, private non-profit, and private for-profit institutions of higher education in the United States that award bachelors, masters, or doctoral degrees.
Methods:
Using digital content analysis, human coders analyzed institution websites for informational text indicating student telemedicine services.
Results:
Findings indicate that a minority of 4-year and above institutions offer telemedicine access. Institution type, institution size, and the presence of campus student health services were predictive. Endowment size and Minority Serving Institution status were not predictive.
Conclusion:
This study illustrates the ongoing need for increased access to remote health services across higher education, especially among smaller private and public non-profit colleges and universities and all private for-profit institutions.
Keywords: telemedicine, student health, COVID-19
Introduction
In April 2020 over 1,000 college and university campuses in the United States closed due to COVID-19. This impeded in-person access for twenty-one million students who depended on campus-based student health and counseling services as their primary healthcare provider. Almost overnight, remote telemedicine access in higher education became a necessary solution and an understudied phenomenon.1
A survey of 257 campuses by the American College Health Association in August 2020 provides a benchmark assessment of the rate of provision of telemedicine services for students. The survey indicated that 66.5% of primarily urban, 4-year, predominantly white institutions (PWIs) reported plans to provide medical care via telemedicine in the fall 2020 semester.2 However, the undersampling of Historically Black Colleges and Universities (HBCUs) and Tribal Colleges and Universities (TCUs), as well as private for-profit colleges and non-urban PWIs limits the generalizability of the findings across the higher education landscape more broadly.
We sought to characterize telemedicine access in early 2021 and identify predictive factors in higher educational provision of remote healthcare. We followed the ACHA definition of telemedicine services as remote online access to clinical medical care from physicians, nurse practitioners, physician assistants, and nurses. This excludes telepsychiatry and counseling and mental health services, as well as health promotion/wellness services, which the ACHA measured separately.2
We hypothesized that a minority of 4-year institutions offered telemedicine access. We further hypothesized that institution size, degree of urbanicity, presence of a campus student health center, endowment size, and non-profit status would positively correlate with telemedicine provision. We also hypothesized that 4-year PWIs would have greater rates of telemedicine access than 4-year HBCUs and 4-year TCUs.
Materials and Methods
Using the National Center for Education Statistics (NCES) database, we identified 6,374 institutions of higher education in the U.S. in 2019. Institutions in the following categories were excluded from the study: international universities with satellite campuses or online-only programs in the United States, trade schools, non-accredited, and non-degree-granting institutions, and institutions at which the highest level degree awarded is an associate’s degree. The remaining 2,842 colleges and universities are “4-year” institutions and above, defined as accredited public non-profit, private non-profit, and private for-profit institutions of higher education in the United States that award bachelors, masters, or doctoral degrees, as recognized by the NCES. (Figure 1)
Figure 1.
The selection and exclusion of higher education institutions (NCES).
The study followed social scientific protocols for digital content analysis in the field of health research.3-5 We searched for two medium-specific and salient formal message units: (1) informational text and/or hyperlinked text on the website indicating the presence of a student health center on campus and (2) informational text and/or hyperlinked text on the website indicating student access to telemedicine services.
Human coders used a simple Google search of the institution name and key search terms to resemble a generic student query about health services available to enrolled students (e.g., “Norfolk State University” AND “student health center”; “Norfolk State University” AND “student health services”; “Norfolk State University” AND “student medical services”) and telemedicine services available to enrolled students (e.g., “Norfolk State University” AND “student telemedicine”; “Norfolk State University” AND “student telehealth”; “Norfolk State University” AND “student health online medical services”). In pilot testing, we found searches from Google’s homepage to be more efficient at finding links to student health websites from specific schools than using the embedded search bars from the website (.edu) of an individual school. We assessed video-based, phone-based, and store-and-forward remote medical services. Data collection did not involve human subjects and adhered to Duke University IRB standards and in correspondence the IRB said it thus did not require review. Definitions of additional terms are included in Appendix 1.
Coders were trained in a standardized content analysis protocol for the binary coding scheme and an explanation of how to complete the coding form document. Coders were tested for reliability with sample schools. Each website was coded by two researchers with staggered overlap, to ensure intercoder reliability. We performed kappa testing to report intercoder reliability. Written instructions given to coders are included in Appendix 2.
Responses were coded in REDCap between November 2020 and January 2021. In cases where there was a discrepancy between two coders, one study author (AH) repeated coding and analysis. Two authors (AH and CH) then reviewed the reason for the discrepancy and made a final determination. Two institutions with identical names but separate NCES Institution IDs were excluded from the general pool and coded directly by two study authors (AH and CH).
We developed multivariate logistic regression models to better understand predictors of telemedicine provision. Modeling included all categorical institutional characteristics as well as endowment size, financial assets, graduation rate and retention rate for 2019 as reported by NCES. We were especially interested in how the financial position of an institution or historic academic performance of students might relate to provision of telemedicine and included log-adjusted endowment and graduation rates in a final model. Log-transformation was performed on endowment statistics as the data were highly skewed. We also included HBCU status in models under evaluation but found it to be highly multi-collinear with most of the other variables, so it was removed.
Results
The study sample (n=364) included the complete substrata of 4-year HBCUs (n=88) and 4-year TCUs (n=16) and a stratified random sample (n=260) of the remaining 2,738 4-year institutions on the basis of Institutional Locale and Institutional Size. (Figure 2.) The majority of the sample were four-year colleges (n=269, 72% of the sample) or graduate-only programs (n=34, 11% of the sample). The sample also included some four-year colleges with at least one baccalaureate degree program that award more than 50 percent of degrees at the associate's level (n=61, 16% of the sample). (Table 1.)
Figure 2.
Sample substrata (NCES).
Table 1.
Summary statistics for telemedicine and its potential correlates.
| Offers Telemedicine | Does not offer Telemedicine |
Total | Adjusted F Test |
|
|---|---|---|---|---|
| Institutional category | n=131(36.80%) | n=233(63.20%) | n=364 | 27.01 (p<0.0001) |
| Not primarily baccalaureate or above (3) | 4 (7.2%) | 57 (92.8%) | 61(16.24%) | |
| Primarily baccalaureate or above (2) | 123 (47.79%) | 146 (52.21%) | 269(72.19%) | |
| Graduate degrees only (1) | 4(9.9%) | 30(90.10%) | 34(11.58) | |
| Sector of institution | 6.89 (p=0.009) Public not-for-profit vs Private not-for-profit only | |||
| Public, not-for-profit (1) | 61(53.16%) | 65 (46.84%) | 126(29.29%) | |
| Private, not-for-profit (2) | 70(35.61%) | 138(64.39%) | 208(59.62%) | |
| Private, for-profit (3)* | 0 | 30(100.00%) | 30(11.09%) | |
| Degree of urbanization | 1.75 (p=0.1565) | |||
| 1- City | 73(40.71%) | 112(59.29%) | 185(48.78%) | |
| 2- Suburb | 22(26.56%) | 60(73.44%) | 82(27.33%) | |
| 3- Town | 29(47.75%) | 37(52.35%) | 66(18.75%) | |
| 4- Rural | 7(30.93%) | 24(69.07%) | 31(5.14%) | |
| Institution size | 14.92 (p<0.0001) | |||
| Under 1000 students | 19(13.06%) | 122(86.74%) | 141(38.50%) | |
| 1000-4999 students | 57(43.20%) | 83(56.80%) | 140(36.46%) | |
| 5000-9,999 students | 26(55.05%) | 16(44.95%) | 42(10.21%) | |
| 10,000-19,999 students | 16(62.30%) | 7(37.70%) | 23(8.17%) | |
| >20,000 students | 13(72.22%) | 5(27.78%) | 18(6.66%) | |
| Land grant Status | 0.1742 (p=0.6766) | |||
| Land grant-inst. (1) | 14(42.60%) | 25(57.40%) | 39(2.77%) | |
| Not a land-grant inst. (2) | 117(36.62%) | 208(63.38%) | 325(97.23%) | |
| HBCU/TCU institution status | ||||
| HBCU/C | 35(39.69%) | 53(60.31%) | 88(3.25%) | 0.2234 (p=0.6367) HBCU vs predominantly white institutions |
| TCU/C* | 0 | 16(100.00%) | 16(0.59%) | |
| Predominantly white inst. | 96(36.92%) | 164(63.08%) | 260(96.15%) | |
| Student Health Services | 211.61 P<0.0001 | |||
| Provides student health services | 128(66.72%) | 90(33.28%) | 218(53.99%) | |
| Does not provide student health services | 3(0.20%) | 143(99.80%) | 146(46.01%) |
We had high intercoder reliability for the presence of a student health center (Kappa= 0.9893, se= 0.0078, 95% CI 0.9739, 1.000, Accuracy= 99.47%) and intercoder reliability on access to telemedicine services (Kappa= 0.7801, se= 0.0395, 95% CI 0.7023, 0.8579), Accuracy=89.98%. To compare institutional characteristics associated with telemedicine provision, we used design-adjusted Wald F tests for categorical predictors with jackknife adjustment method for standard errors to account for design effects, implemented using PROC SURVEYFREQ in SAS.
A minority of institutions in the sample offered telemedicine by late 2020 (131/364, 37%). (Table 1.) Institutional category was a strong predictor of telemedicine provision (p<0.001). Almost half of institutions awarding primarily baccalaureate degrees offered telemedicine (123/269, 48%). In contrast, few institutions awarding graduate degrees only offered telemedicine (4/34, 10%) and few institutions awarding primarily non-baccalaureate degrees provided telemedicine (4/61, 7%).
Institution sector was also highly predictive of telemedicine provision. Of 30 private, for-profit institutions in the sample, none offered telemedicine (0/30, 0%). A higher proportion of public, not-for-profit institutions provided telemedicine (61/126, 53%) than private, not for profit institutions (70/208, 36%) (p=0.009).
Schools in the sample that had student health services (128/218, 67%) were more likely to offer telemedicine than those without student health services (3/146, 0.2%). Larger institutions were also more likely to offer telemedicine in bivariate comparisons (p<0.001). Telemedicine provision was not significantly associated with degree of urbanicity (p=0.16), Land-Grant Institution (LGI) status (p=0.68) or HBCU status as compared to PWIs (p=0.64). None of the TCUs offered telemedicine (0/16, 0%), though many referred students to nearby Indian Health Service facilities for healthcare needs.
In multivariate analyses, public, not-for-profit institutions had increased odds of offering telemedicine (aOR 3.32 95% CI 1.43-7.72) as compared to private, not-for-profit institutions. (Table 2.) Institutions offering primarily baccalaureate degrees or above also had much higher odds of telemedicine provision (aOR 8.86 95% CI 1.9-41.1) as compared to those offering not primarily baccalaureate or above. Larger log endowment size did increase odds of telemedicine provision (aOR 1.31 95% CI 1.0-1.7), while graduation rate was not a significant predictor (aOR 1.02 95% CI 0.998-1.05). (Table 2.)
Table 2.
Factors associated with availability of telemedicine services: adjusted odds ratios.
| Variable | aOR | 95% Confidence Limits |
ChiSq- value |
Pr > ChiSq | |
|---|---|---|---|---|---|
| Sector of institution | |||||
| public non-profit vs private non-profit | 3.317 | 1.426 | 7.717 | 7.84 | 0.0057 |
| Institutional category | |||||
| Primarily baccalaureate or above vs Not Primarily baccalaureate or above | 8.861 | 1.910 | 41.108 | 7.82 | 0.0056 |
| Endowment | 1.307 | 1.006 | 1.699 | 4.05 | 0.0453 |
| Graduation rate | 1.022 | 0.998 | 1.046 | 3.23 | 0.0737 |
Discussion
A minority of 4-year institutions in the sample offered remote telemedicine access, supporting our hypothesis. This demonstrates a need for increased access to services across the higher education sector. Unsurprisingly, telemedicine provision was correlated with existing student health infrastructure, in the form of on-campus student health centers, and higher institution resources, measured by endowment size.
The ACHA survey from August 2020 provided 66.5% as a benchmark rate of provision of telemedicine services for students at primarily urban, 4-year PWIs.2 Our study provides a more diverse sample and a clearer picture of sectors that are providing comparable rates of service, including HBCUs and larger universities, as well as sectors that are not, such as private for-profit colleges and smaller, private non-profit colleges.
Our results suggest a number of institutional features that correlate with telemedicine access. Remarkably, all 30 private, for-profit institutions do not offer telemedicine. This lack of provision stands in stark contrast to the overall availability of services within higher education and may support recent critical analysis of for-profit colleges as exacerbators of social inequality.6,7
Minority-serving institution status was not predictive, contrary to our hypothesis. HBCUs are underfunded and have two-thirds the revenue per enrolled student compared to colleges in general.8 Additionally, there are no HBCUs among the 100 schools in the U.S. with endowments above $1 billion.8 As endowment size was positively correlated with telemedicine provision, this suggests that HBCUs are finding ways to provide telemedicine access for students at similar rates to schools with higher relative wealth position within higher education.
The case of 4-year TCUs is inconclusive. Zero of the 16 TCUs analyzed offered telemedicine. Anecdotally, most TCU institution websites directed students to local clinics provided by the Indian Health Service, a federal health program for American Indians and Alaskan Natives (AI/AN). According to the U.S. Department of Education, AI/AN students comprise 78 percent of the combined total enrollment at TCUs.9 Additional research is needed to understand whether AI/AN students utilize Indian Health Service clinics, including telemedicine services, and where the 22 percent of non-AI/non-AN students at TCUs obtain healthcare.
The presence of an on-campus student health center was predictive of telemedicine services, confirming our hypothesis. This reinforces the importance of on-campus health centers and suggests increased health vulnerability for students attending campuses at the one in six 4-year colleges and universities without student health services.10
Our study is not without limitations. We evaluated availability of telemedicine on the institution rather than student level, and the 37% of institutions offering telemedicine serve greater than 37% percent of students. For example, we did find a high proportion of the largest campuses (71% of those with over 10,000 students) offered telemedicine. While 13 percent of campuses enrolled 10,000 or more students in 2017, they accounted for 61 percent of total college enrollment.11 In contrast, 42 percent of institutions had fewer than 1,000 students; and enrolled only 3 percent of all higher education students.11 This offers some reassurance that the proportion of students with telemedicine access is higher than the proportion of campuses offering telemedicine. Content analysis offers only a “snapsot” understanding of website information, meaning that our analysis is true of only the time that the websites were coded. Following ACHA definitions, this study excluded telepsychiatry and telecounseling services from assessment of telemedicine services, which limits conclusions about the availability of mental health services during the pandemic. 2 The study indicates only the presence or absence of student health centers and telemedicine services, which limits conclusions about the quality and accessibility or comprehensiveness of care. Most studies of campus-based student health utilize self-reported institutional offerings through survey methods. While this method guarantees validity, it threatens to undermine generalizability due to preferential sampling bias. Visual content analysis offers a strong claim to generalizability through the inclusion of underrepresented institutions and stratified random sampling with NCES institutions. One threat to validity is the possibility that the message unit regarding telemedicine services is behind a student login portal.
Our findings indicate an ongoing need for increased access to services across the higher education sector, especially among smaller public non-profit and private non-profit colleges and universities. Findings also raise significant concerns about health access for students at private for-profit universities. Future research should investigate rates of telemedicine access during the pandemic relative to rates of higher educational attainment.
Acknowledgements
Special thanks are due to the 2020-2021 Bass Connections team from Duke University, including Emily Chen, Lauren Holt, Becca Lane, Emily Sen, Sydney Morrow, Rachel Proudman, and Lucy Zheng.
Declaration of Interest
The authors have no conflicts of interest to report. The authors confirm that the research presented in this article met the ethical guidelines, including adherence to the legal requirements, of the United States of America and the Institutional Review Board of Duke University. Jonas Swartz is a Women’s Reproductive Health Research Scholar (K12HD10383-01)
Appendix 1
Content Analysis Terms
Descriptive qualitative content analysis is a growing area of research in health and gender.3,5 Manganello and Blake 2010 found a steady increase of content analysis research in interdisciplinary health literature between 1985 and 2005.12 Content analysis is also emerging in medical fields, such as nursing, psychiatry, and pediatrics.13
This study follows social scientific protocols for content analysis.3-5 It pursues descriptive qualitative content analysis of college and university student health websites to determine the numerical frequency of telemedicine information and services. The key analysis is a count of college student health websites with information and links to telemedicine services.
College and university websites are defined as official internet landing websites where the website url ends in .edu. Google results or university website links that directed students to non-university health provision websites like www.firststudent.com, and www.coverage2u.com were excluded from the study. In two instances, a university-run website (.edu url) directed users to a university-run health center website with a .com url. These two schools were included in the study and coded Yes for provision of a campus student health center.
Some student health clinics contracted with insurance providers to provide telemedicine access for students with certain insurance providers. For instance, the HealthistYou App offers 24/7 access to doctors through Teladoc for students on UnitedHealthcare plans, and6www.studentbluenc.com offers telemedicine to students on BlueCross and Blue Shield of North Carolina plans. Although these programs are advertised as student-specific, they are not provided by the college or university health center providers and the university website redirects students to non-university telemedicine providers. Most importantly, these contracted services often exclude students with Medicaid coverage, even if they are enrolled full-time. Because these telemedicine services are not available to all enrolled students, they were coded “no" for telemedicine services.
A student health center describes a physical building, suite, or room on campus with a university-employed healthcare provider: nurse practitioner, medical doctor, or registered nurse. Mental health and counseling facilities were excluded from the study. Fitness centers, exercise facilities, and “wellness” centers that offer nutritional or diet counseling were excluded. Referrals to local county- or city-based health clinics were excluded.
Informational text is conceptualized as informational text embedded with an electronic link that directs users to another site or place within the electronic document.
Hyperlinks and hyperlinked text are “inscriptions of communicative and strategic choices on the part of site producers,” which, for the purposes of this study, refers to the college and university site producers.14
Telemedicine describes the delivery of health care services where distance between the patient and the provider is “a critical factor,” and where information and communication technologies are utilized to facilitate the necessary information for “diagnosis, treatment and prevention of disease and injuries, research and evaluation, and for the continuing education of health care providers, all in the interests of advancing the health of individuals and their communities.”15 While telemedicine is increasingly understood as video-based medical services, services by phone were included in this study.
There are 101 Historically Black Colleges and Universities (HBCUs) in the United States, of which 88 met study criteria of accredited public, private, and for-profit institutions of higher education in the United States that award bachelors, masters, or doctoral degrees, as recognized by the National Center for Education Statistics.
There are 32 Tribal Colleges and Universities (TCUs) in the United States, of which 16 met study criteria of accredited public, private, and for-profit institutions of higher education in the United States that award bachelors, masters, or doctoral degrees, as recognized by the National Center for Education Statistics.
Appendix 2
Instructions to Coders
The RedCap coding scheme included the following components:
External Demographics (According to NCES database)
Institution ID
Institution Name
Special Mission
Institutional Locale
Institutional Size
Internal Reliability
-
6.
Coder ID
-
7.
Date of Coding
Message Units
-
8.
Unit 1 Presence
-
9.
Unit 1 Message Unit, if applicable
-
10.
Unit 1 URL, if applicable
-
11.
Unit 2 Presence
-
12.
Unit 2 Message Unit, if applicable
-
13.
Unit 2 URL, if applicable[SJ13]
Coder instructions for Message Unit 1
1. Search in Google: "institution name" and "student health center."
2. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating the presence of a student health center. If the result is determinate of a message unit, complete the record log "Yes", input the message unit and URL. If the result is indeterminate, continue to step 3.
3. Search in Google: "institution name" and "student health services."
4. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating the presence of a student health center. If the result is determinate of a message unit, complete the record log "Yes", input the message unit and URL. If the result is indeterminate, continue to step 5.
5. Search in Google: "institution name" and "student medical services."
6. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating the presence of a student health center. If the result is determinate of a message unit, complete the record log "Yes", input the message unit and URL. If the result is indeterminate of a message unit, complete the record log "No".
Coder instructions for Message Unit 2
1. Search in Google: "institution name" and "student telemedicine".
2. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating student access to telemedicine services. If the results are determinate of a message unit, complete the record log "Yes," input the informational text and website URL. If the results are indeterminate, continue to step 3.
3. Search in Google: "institution name" and "student telehealth".
4. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating student access to telemedicine services. If the results are determinate of a message unit, complete the record log "Yes," input the informational text and website URL. If the results are indeterminate, continue to step 5.
5. Search in Google: "institution name" and "student health online medical services".
6. Based on search results, evaluate the presence of informational text and/or hyperlinked text on the website indicating student access to telemedicine services. If the results are determinate of a message unit, complete the record log "Yes," input the informational text and website URL. If the results are indeterminate, complete the record log "No".
References
- 1.Hollowell A, Swartz JJ, Proudman R. Telemedicine access and higher educational attainment. J Am Coll Health. 2021;online. 10.1080/07448481.2021.1891085?journalCode=vach20 [DOI] [PubMed] [Google Scholar]
- 2.ACHA. The COVID-19 Pandemic’s Effect on Campus Health and Well-Being Services. August 4-7, 2020. American College Health Association; 2020. Accessed December 10, 2020. https://www.acha.org/documents/Resources/COVID_19/COVID-19_Effect_On_Campus_Health_Services_REPORT_3_August_2020.pdf [Google Scholar]
- 3.Neuendorf KA, Neuendorf KA. Content Analysis—A Methodological Primer for Gender Research. Sex Roles. 2011;64(3):276–289. doi: 10.1007/s11199-010-9893-0 [DOI] [Google Scholar]
- 4.Jordan A, Kunkel D, Manganello J, Fishbein M, eds. Media Messages and Public Health: A Decisions Approach to Content Analysis. 1st Edition. Routledge; 2008. [Google Scholar]
- 5.Neuendorf KA. The Content Analysis Guidebook. 2nd ed. SAGE Publications, Inc; 2019. [Google Scholar]
- 6.Cottom TM. For-Profit Colleges Thrive Off of Inequality. The Atlantic. Published February 22, 2017. Accessed July 1, 2020. https://www.theatlantic.com/education/archive/2017/02/the-coded-language-of-for-profit-colleges/516810/ [Google Scholar]
- 7.Cottom TM, Kelton S. Lower Ed: The Troubling Rise of For-Profit Colleges in the New Economy. Reprint edition. The New Press; 2018. [Google Scholar]
- 8.Startz D. When it comes to student success, HBCUs do more with less. Brookings. Published January 18, 2021. Accessed May 18, 2021. https://www.brookings.edu/blog/brown-center-chalkboard/2021/01/18/when-it-comes-to-student-success-hbcus-do-more-with-less/ [Google Scholar]
- 9.Department of Education. Tribal Colleges and Universities ∣ White House Initiative on American Indian and Alaska Native Education. Published 2021. Accessed April 8, 2021. https://sites.ed.gov/whiaiane/tribes-tcus/tribal-colleges-and-universities/
- 10.Lemly DC, Lawlor K, Scherer EA, Kelemen S, Weitzman ER. College Health Service Capacity to Support Youth With Chronic Medical Conditions. Pediatr Evanst. 2014;134(5):885–891. doi: 10.1542/peds.2014-1304 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Digest of Education Statistics. National Center of Education Statistics; 2018. Accessed May 18, 2021. https://nces.ed.gov/programs/digest/d18/ch_3.asp [Google Scholar]
- 12.Manganello J, Blake N. A Study of Quantitative Content Analysis of Health Messages in U.S. Media From 1985 to 2005. Health Commun. 2010;25(5):387–396. doi: 10.1080/10410236.2010.483333 [DOI] [PubMed] [Google Scholar]
- 13.Neuendorf KA. Reliability for content analysis. In: Jordan A, Kunkel D, Manganello J, Fishbein M, eds. Media Messages and Public Health: A Decisions Approach to Content Analysis. 1st Edition. Routledge; 2008:67–87. [Google Scholar]
- 14.Foot K, Schneider SM, Dougherty M, Xenos M, Larsen E. Analyzing Linking Practices: Candidate Sites in the 2002 US Electoral Web Sphere. J Comput-Mediat Commun. 2003;8(4):0–0. doi: 10.1111/j.1083-6101.2003.tb00220.x [DOI] [Google Scholar]
- 15.WHO, ed. Telemedicine: Opportunities and Developments in Member States: Report on the Second Global Survey on EHealth. World Health Organization; 2010. [Google Scholar]