OUP user menu

★ Research Paper ★

Patient Internet Use for Health Information at Three Urban Primary Care Clinics

Suzanne Dickerson DNS, RN, Amber M. Reinhart MA, Thomas Hugh Feeley PhD, Rakesh Bidani MD, Ellen Rich MD, Vinod K. Garg MD, Charles O. Hershey MD
DOI: http://dx.doi.org/10.1197/jamia.M1460 499-504 First published online: 1 November 2004


Objective: To survey a cross section of patients presenting to three urban primary care clinics to understand online health information search behaviors.

Design and analysis: At three urban primary care clinics affiliated with University at Buffalo, School of Medicine, 315 patients were interviewed. Interview questions included items on education, demographic information, employment, number of current prescriptions, insurance, online access, and specifics of health-searching behaviors. Chart review determined patient body mass index and number of chronic illnesses. Logistic regression and χ2 statistics were used to investigate the relationship between patient characteristics and the proportion of patients who use the Web for seeking health information.

Results: Approximately 53% of respondents reported using Web or e-mail in the past year and 68% (33% of total sample) of those who accessed the Web used it to search for health information. The two most commonly cited search areas included information about a physical illness and nutrition/fitness. Education and race significantly predicted online health-seeking behavior when considering all factors in the study. Many patients (22%) relied on friends and family to navigate the Web, and 45% of patients reported that the information that they sought was unrelated to their clinical visit.

Conclusion: Current use of the Internet for health information was limited among more disadvantaged patient groups. More research is needed to examine the relationship between health-seeking behavior and patients' management of their health and well-being.


Computer-literate patients are seeking to take more control of their health by using the Internet to seek out information related to illness and treatment.13 The information provided on the Web helps patients to better manage illness and to make informed choices, and ultimately changes the dynamics of the patient–provider relationship. For good reason, health care providers have a healthy skepticism of the quality of the information readily available on the Internet.47

Despite concerns about the integrity of the health information on the net, Americans are increasingly using the Web for information retrieval and personal communication to assist in understanding and managing illness. The National Telecommunications and Information Administration8 reports indicate that in 2000, 44% of individuals had access to the Internet; this percentage increased to 54% in 20029 and is projected to exceed 80% by 2005.10 In the United States, the number of Internet users is estimated to be between 122 million7 and 166 million.11 Although Internet use is on the rise overall, there is evidence that it is also increasing in particular demographic populations including the poor, the less educated, the elderly, and minority groups.7

Internet Use and Health-seeking Behavior

A primary goal of the current paper is to identify the proportion of patients at three primary care ambulatory clinics who use the Internet for health-seeking purposes. Health seekers are defined as Internet users who search the Web for information on health topics.7 The percentage of patients who use the Internet regularly is still less than clear. Recent estimates range from 20% access to 47%7,1214 access, whereas one estimate of the general public's access is as high as 60%.7 In an earlier study of patient access conducted in 1997 and 1998, Robinson et al.15 found 20% of primarily African-American patients to have access and only 18% reported owning a computer. In a cross-sectional study of Veterans Affairs patients presenting to seven general internal medicine clinics, 29% of patients from June 1999 to March 2000 reported that they used or planned to use the Internet.13 Two other studies reported 37% access for lung cancer patients16 and 47% for cancer patients in Hamilton, Ontario, Canada. Thus, it may be concluded that patients who do have Web access may be in the minority.

How many patients use the Internet for searching resources for health information? The recent Pew Internet and American Life Project7 randomly interviewed 2,038 Americans in November and December of 2002 and estimated that more than 80% of Internet users, or approximately 93 million Americans, have searched for health information online in the past. The authors of the report conclude that “this makes the act of looking for health or medical information one of the most popular activities online, after email (93%) and researching a product or service before buying it (83%).”

Two considerations qualify the conclusions of the Pew survey findings. First, the Pew study asked whether respondents have “ever” searched for health information, whereas two recent studies12,17 limited the scope of the question to “the past year” or health seeking related to the current medical appointment. Second, recent data by Baker et al.12 collected in 2001 and January 2002 indicate that respondents who report health-seeking behavior in the past year indicate that going online to search for health information is seldom done; 78% of health seekers reported going online only once every 2 to 3 months for health information. It is important to note that the Baker et al. sample consisted of individuals who already have Web access.

There are clear trends in characteristics of individuals who use the Internet in general and those who use the Internet for health seeking. Specifically, online users are more likely better educated, younger, female, and European American and earn a higher income.7,1217 Baker et al.12 also found that patients with fair to poor health (compared with those with good or excellent health) were more likely to be health seekers.

In the current study, a cross section of patients presenting to their primary care physician was interviewed. The current sample of patients, it is expected, represents a diverse population in terms of race, ability to pay, age, and health status. Thus, the ability to generalize results to primary care patients across urban clinics is of great importance. Two research questions guided our study protocol:RQ1: Using an urban sample of patients, what is the percentage of patients who regularly use the Internet and how many of those with access use the Internet for health-related information?RQ2: Considering the sample of patients who use the Internet for health information, how often do they use the Web, how do they search for health information, and how useful is the information in managing their or another's illness?

In attempting to answer these research questions, a second goal is to investigate whether there exists a “digital divide” in our patient population. That is, the current study will attempt to quantify background characteristics of patients that differentiate those with Internet access and health-seeking experience and those who report no online experience.

The current study promises to add much to our knowledge of patient use of the Internet for health-seeking reasons. First, patients will be asked to report whether they have searched online in the past year and whether they regularly use the Internet. Two recent studies7,13 used alternative question wording that may have provided misleading results. Second, several questions will probe patients' motivation for health seeking and whether patients found the exercise useful and informative. In so doing, these data may provide information linking patients' health-seeking tendencies to the purpose of the clinic visit, especially if patients self-report searching for information just before their clinic visit.


Setting and Participants

Three university primary care practice sites located in urban Buffalo, NY, were used in this study. All three sites are staffed with clinical faculty members affiliated with the Department of General Internal Medicine in the School of Medicine and Biomedical Sciences at the State University of New York at Buffalo. Academic Medical Services (AMS) was the first site used; AMS is a private university practice site and provides clinical care for primarily working insured and retired patients. The Primary Health Care Clinic (PHC) is a primary care clinic for the Erie County Medical Center, one of two primary university training hospitals. A third setting for the study was the Matthew J. Gajewski Clinic (MJG), a public clinic servicing mainly poor, uninsured patients on the east side of Buffalo. AMS and PHC are located in health professional shortage areas and medically underserved areas, according to the U.S. Department of Health and Human Services.18

The initial study design selected patients using a random numbers table with each day's patient schedule as a template. However, due to the high no-show rate of patients at the MJG site, a decision was made to interview all patients at MJG for the data collection period. Thus, a consecutive sampling technique was employed at the MJG site and a daily random sample of patients was used at the AMS and PHC sites. Demographic information for patients at the PHC site were compared with previously reported19 demographic information to determine generalizability; the results showed a similar patient profile in terms of gender, race, insurance coverage, and number of chronic illnesses.

Patient charts were reviewed with attending physicians (RB, ER, VKG) to prescreen patients for study eligibility and to record patient information related to number of chronic diagnoses and information on patient height/weight. All interviewers received institutional review board and Health Insurance Portability and Accountability Act training before data collection. Patients were excluded if they were minors, prisoners, or institutionalized; could not speak English; had a neurological disorder or a severe psychiatric condition; or were unable to give consent for various reasons. Of the 342 patients considered for inclusion, 315 participated, for a 92% participation rate. Eleven patients refused to participate, 12 were new patients, one patient could not speak English, and three patients had dementia. One hundred forty-six patients were interviewed at AMS, 122 at PHC, and 47 patients at MJG; the difference in sample size was a function of patient volume and the no-show rate of patients at each clinic for the study time period.


After the attending physician determined patient eligibility, patients were escorted into the examining room by the nursing staff at the individual site. Nurses conducted their typical examination depending on the nature of the visit and then left the examining room. Upon completion of the nurse's consultation, one of three female research assistants entered the examining room while the patient waited for the attending physician. Patients were informed of the nature of the study, invited to participate, and asked for their signed consent. This proved to be an ideal time to approach patients because they typically wait some time before being examined by the physician. The interviews were conducted during a 7-week period between June 9, 2003, and July 25, 2003. Each interview lasted, on average, 5 to 10 minutes. After patient interviews, participants were thanked for their time and given an information sheet providing Web addresses of reputable health information organizations. The Social and Behavioral Sciences Institutional Review Board at the State University of New York at Buffalo approved the survey instrument and study protocol.

Survey Instrument

All surveys were completed by the three research assistants who interviewed each patient participant and asked questions about background information (e.g., education, language, employment, demographics), number of current prescriptions, and questions about Web access.

Demographics and Health Status

The number of patient diagnoses was taken from individual patient charts as were height and weight at the time of visit. Body mass index was used to determine patient obesity from weight and height information. An equal percentage (47%) of patients across the clinics were black (n = 149) and white (n = 147); 18 patients were Hispanic, Asian, or other.

Access to Internet

Patients were asked to report their level of access to the Web and their use of the Web for health information. Specifically, patients were asked, “Have you used information from either the Web or e-mail within the past year? (yes/no).” Patients were also asked to report how often they go online (more than once per day/once per day/more than once per week/more than once per month/less often) and how they access the Web (e.g., own personal computer, library, work).

After obtaining general Web access information, patients were asked individual questions about their use of the Web to obtain health-related information for themselves or for others and/or if they used the Web to e-mail a health care provider. The types of medical Web sites accessed and the usefulness of the information were also interview questions for patients who self-reported using the Web for health information.

Study participants were also asked what type of information they looked up on the Web and how they found out about the Web page or information source (i.e., Web search, news article, friend). Patients were asked about the timing of the Web search (i.e., before or after the clinic visit), whether they learned something new (yes/no), and whether they spoke to their health care professional about the information (yes/no). Finally, patients were asked whether the health information obtained via the Web affected their clinic visit, treatment, or the manner in which they cope with illness and whether it led the patient to ask new questions or seek second opinions.

Reponses were analyzed using standard tabulations. Chi-square test was used at the bivariate level to explore differences between patient characteristics and online access in general and online access for health information. Logistic regression analysis was used to predict health searching from employment, education, race, gender, insurance, age, number of prescriptions, and number of chronic illnesses. Analyses were performed using SPSS 11.5 and 0.05 statistical significance was used unless otherwise noted.


Patient Demographics

The average patient age was 53.76 years old (SD = 16.03), and 57% of patients were female. Twenty-four percent of patients did not graduate from high school, 31% had a high school diploma, 23% attended college, and another 21% earned college degrees. One hundred sixteen patients were currently working, 43 in-between jobs, 89 retired, and 64 disabled. Three did not respond to this question. All other demographic and background information can be found in Table 1.

View this table:
Table 1

Online Access for Health-Related Use by Patient Characteristics

Patient CharacteristicsnProportion Who Used WebProportion Who Used Web for Health
Did not attend college*1750.350.18
Attended college1370.760.50
White or other*1650.620.37
Not employed*1960.440.27
   ≥75 yr*,390.260.21
   65–74 yr400.280.18
   50–64 yr1040.530.31
   35–49 yr970.660.43
   18–34 yr340.760.38
No. of chronic conditions*
No. of current prescriptions
Commercial insurance1700.620.38
BMI category
Morbidly obese330.580.36
Clinical site

AMS = Academic Medical Services; BMI = body mass index; MJG = Mathew J. Gajewski clinic; PHC = Primary Health Care Clinic.

  • * χ 2 test of proportions significant at 0.05 level for using the Web.

  • χ 2 test of proportions significant at 0.05 level for using the Web for health purposes.

Health Status

Table 2 lists the number of patients with particular diagnoses on their medical chart. The three most common diagnoses included hypertension (49%), arthritis (22%), and diabetes (21%). Each patient had, on average, 4.32 diagnoses and self-reported current use of 5.10 prescriptions. Fifty-four percent of patients had commercial insurance. Ninety-three patients (29%) were normal weight or underweight, 81 patients (26%) were overweight, and 130 patients (41%) were obese or morbidly obese.

View this table:
Table 2

Patient Health Status

DiagnosisNo. of Patients (% of Sample)
Hypertension154 (49)
Degenerative joint disease/arthritis69 (22)
Diabetes67 (21)
Mental illness55 (17)
Smoking48 (15)
High cholesterol48 (15)
Obesity43 (14)
Coronary artery disease37 (12)
Substance abuse23 (7)
Pulmonary disease15 (5)

Use of Internet

Overall, 166 patients or 53% of sample reported using the Web or e-mail in the past year. Almost one-half of Web users (48%) go online at least once per day, 20% more than once per week, and 30% more than once per month or less often. Patients (153 of 166 with access) were asked how they accessed the Web, and 70% reported doing so using their home computer or a computer at the public library or work. Almost 30% did not access the Web themselves; instead, patients reported accessing the Web through a family member, neighbor, or friend. Table 1 lists the percentage of patients who access the Internet by patient characteristics. A higher proportion of patients reported online access who were educated (i.e., attended college), employed, younger, diagnosed with fewer chronic illnesses, commercially or privately insured, and patients at the AMS site. Blacks reported less online access compared with the rest of the study sample.

Use of the Internet for Health Information

Of the 166 persons who use the Internet, 103 reported obtaining health information from the Web (33% of the total sample). The majority of patients (68%) who accessed the Web for health-related information did so by performing a Web search and accessed information (73%) from a medical site providing general information. Another 46% of health Web users accessed a Web site for a specific disease or problem. Table 3 outlines the information that patients described looking up on the Web. The most frequently cited motivation for searching the Web for health information was to find out more information about a physical illness (n = 82). Eighty-seven percent of those who access the Web for health information found the information either somewhat or very useful, and 85% “learned something new.” Few patients (13%) searched for information before the visit to their primary care doctor, another 43% accessed health information after the clinic visit, and 45% searched for information unrelated to their clinic visit.

View this table:
Table 3

Health Information Accessed via the Web by Patient Sample

Information AssessedFrequency (% of Patients Using Web for Health)
Information about a physical illness82 (80)
Information about nutrition and fitness59 (57)
Information about specific doctors, hospitals, or medicines37 (36)
Information on alternative treatments33 (32)
Information about mental health20 (19)
Information on experimental treatments17 (17)
Which company provides the advice15 (15)

Regression results are shown in Table 4 and are the standardized β coefficients, standard errors, and odds ratios (95% confidence interval) for each of the independent factors regressed onto health-seeking behavior. The model explained 17% of the variance in health-seeking behavior (R 2 = 0.17). When controlling for other independent factors, only education and race were statistically significant. The number of chronic conditions and gender were significant at the 0.10 α level. Clearly, education was the single greatest predictor of online search behavior.

View this table:
Table 4

Logistic Regression Results for Predicting Online Health-searching Behaviors

Patient Characteristicsβ CoefficientSE βWald StatisticOR (95% CI)
Employed0.210.320.431.23 (0.66–2.31)
Attended college1.460.3024.344.31 (2.41–7.69)
Black−0.580.294.15*0.56 (0.32–0.98)
Medicare−0.230.510.210.79 (0.30–2.14)
Commercial−0.130.360.130.88 (0.43–1.79)
Age (yr) Female0.480.292.781.62 (0.92–2.86)
   18–340.400.710.331.50 (0.37–5.99)
   35–490.780.591.802.19 (0.70–6.89)
   50–640.190.560.111.21 (0.41–3.60)
   65–74 years old−0.440.630.470.65 (0.19–2.24)
No. of chronic diseases
   00.460.570.641.58 (0.51–4.86)
   1–2−0.720.442.710.49 (0.21–1.15)
   3–4−0.260.370.490.77 (0.37–1.60)
No. of prescriptions
   00.050.490.011.06 (0.40–2.78)
   1–20.590.421.981.81 (0.79–4.12)
   3–40.100.390.061.10 (0.51–2.37)

CI = confidence interval; OR = odds ratio; SE β = standard error of β coefficient. R 2 = 0.17, p < 0.001.

  • * Significance of Wald Statistic at 0.05 criterion.

  • Significance at 0.001 criterion.


The current descriptive study attempted to better understand how patients who present to three urban primary care clinics accessed and used the Internet for health information. As we had anticipated, patient background and health status were quite diverse. However, compared with recent studies in this area,12 the current sample of patients was older and less educated and had a greater number of chronic conditions, on the whole. Consider that the average patient age was 54 years old and the mean number of chronic illnesses, according to patient charts, was 4.32. A study12 conducted in 2001–2002 that found 56% of respondents to report no chronic conditions, and almost 90% of respondents were categorized as being in good or excellent health.

One-third of patients reported using the Internet for health-related information in the past 12 months, and more often patients were investigating information about a specific illness. Whereas in the current study, the patient access (52.6%) and use for health information percentages (61% of users) were less than those of the general population (56% access, 80% use for health information),7 they were still higher than another study of disadvantaged patients.15 Future research should consider the type of sample (i.e., patients versus nonpatients) and sampling method (i.e., random digit dialing versus consecutive sample) to better explain the differences in study findings.

Many of the findings from the current patient interviews are consistent with those reported in the literature. Specifically, those better educated, employed, and younger were more likely to use the Web for health information. A finding that is inconsistent with earlier research12 indicates that patients with a greater number of chronic illnesses were not more likely to search the Web for health information. In fact, the trend in the data presents findings to the contrary, although the odds ratios all failed to reach conventional levels of statistical significance (Table 4), i.e., patients with a greater number of chronic illnesses tended to be less likely to search the Web for health information. This may be partly explained by the significant correlation (r = 0.41) between patient age and number of chronic illnesses. Also worth noting is the finding that patients with commercial insurance were more likely to search the Web for health information. The reasons for this are unknown; however, it may be the case that employers and/or the insurance companies may influence the patients' efforts regarding self-care. Commercial insurance is not a perfect proxy for socioeconomic status in New York because the state and many counties offer partially capitated (commercial) health care plans for residents with particular medical conditions (e.g., addictive disorders, Axis I diagnosis). In addition, the educational level continues to be an important factor regardless of other factors related to Internet use, suggesting that efforts to support education may have wider reaching effects on health promotion.

Patients, on the whole, found the information from the Internet useful and most reported “learning something new.” What is less clear is how online information influenced patient management of their illness and/or the dialogue between the patient and his or her physician. Only 13% of patients searched for information before their visit to their primary care physician, and many patients (45% of those who search for health information) searched for information unrelated to their clinic visit. More in-depth patient interviews, perhaps using focus groups or qualitative interviews, in the future would go far in discovering how different patients use health information in the management of illness.

An interesting result emerged with respect to how patients access the Web. Thirty percent of patients who have Web access do not go online by themselves; instead, these patients rely on salient others to perform Web searches. Moreover, 22% of patient health seekers do their searching via family and friends' access. Together, these data suggest that many patients are willing to expend extra time and energy to better understand information about specific illness(es), health care providers, and general wellness. What is still less than clear is the link between patient health-seeking tendencies and the purpose(s) of the specific clinical visit. Are patients researching health conditions or diagnoses before or after their clinical visit? More research could further identify the critical relationship between health seeking and the clinic visit. Perhaps data on patient satisfaction and adherence could further this line of inquiry.

This study has several limitations that may limit its external validity. First, the sample of 315 patients may be too small to detect independent effects in the multivariate logistic analysis. More power would provide greater confidence and specificity in identifying unique bivariate relations among patient characteristics and health-seeking behavior. A second limitation is the absence of data on patient income levels. Socioeconomic status, it was thought, could be determined by patient health insurance and employment status, but these factors can only be considered crude at best in determining socioeconomic status. Patient income data would have informed whether perhaps the digital divide is narrowing in an urban sample of patients.

The current study also used patient diagnoses as a measure for health status. Perhaps number of diagnoses would better describe “burden of illness” and not necessarily health status because one could be sicker and have only one or two chronic conditions on his or her chart compared with another who is in good health with three or four chronic conditions. An alternative to triangulate patient diagnosis would be to have patients self-report on health status, as in the Baker et al.12 study, or have patients complete the SF-12 or other measures of physical and mental health.

A final consideration is the varying quality of health information available on the World Wide Web.20 Eysenbach et al.20 reviewed 79 distinct studies that evaluate the quality of online health information and conclude that 70% of studies found that quality is a problem. Moreover, studies that draw neutral or positive conclusions use less rigorous criteria to evaluate Web information. Thus, patients may not be receiving accurate, legible, and complete information20 from the few sites that they visit annually.


The current study found that 33% of 315 patients interviewed at three primary care health clinics in Buffalo, NY, reported using the Internet to access information related to health issues. The majority of patients reporting experience with Internet searches related to health did so through a Web search and more often searched information related to a specific illness. When controlling for other factors, only patient education (i.e., if one attended or did not attend college) and race were significant predictors of online health-seeking behavior. Examining a cross section of patients with a greater burden of illness who present to their primary care physician provided a novel vantage point to view how individuals use the Internet in the management or understanding of their health or that of their loved ones.


View Abstract