Keywords

1 Background

Physician information is vital and can make the difference between a positive medical encounter and a negative one when making a physician selection decision. Physician information ranges from general information such as physician’s gender, medical specialty, practice location, education background, to experience-based patients’ rating and reviews on physicians. The availability of patient-oriented online health information and the development of comprehensive health information portals have profoundly changed the way patients seek health information [1–3]. Past studies have confirmed the benefit of accessing physician information for patients’ physician decision-making. Patients can examine the quality reports of health care institutions and patient reviews on physicians prior to the clinical visit [4, 5]. The provision of health care providers’ personal information might enable patient access to physicians’ personal lives that were previously considered beyond the scope of traditional physician-patient relationship [6, 7]. From the patients and caregivers’ perspective, physician information in the form of online physician reviews increase patient empowerment to take proactive actions by supporting useful information on selection of physicians [8, 9]. From the health care providers’ perspective, the patient reviews can be considered as a form of quality evaluation, and improvement can be made based on the review results in order to provide better health care services [10].

Physician information on the Internet can be categorized into three types: professional information, such as curriculum vitae, awards, and publications; personal information, such as social network sites, and family and financial information; and physician rating information that is patient-generated and out of an individual physician’s direct control [11]. However, problems of physician information retrieval remain in multiple aspects. However, certain characteristics of physician information, such as physicians’ bedside manner and personality, are extremely subjective and difficult to describe. In instances when patients and caregivers are looking for information for making decision on physician selection, or when they won’t be able to seeing the same physicians and needing to switch physicians due changes to their health coverage or other reasons, the uncertainty of not knowing which physician to contact and consult increases the difficulty of search for physician information. Consequently, many turn to health-related open forums and solicit recommendations for names of physicians who might be fitting the patients’ health situation and expectation. For example, “I am 37 years old and live in Virginia but currently taking a job assignment in Rome NY for 3 months. I am 8 weeks pregnant with my first child and it’s very difficult to drive back to VA every other week to see my OB at home. I’m so stressed out right now because it’s very important that I’m in OB care during my 1st Trimester but I have to be away from home. I am willing to drive to Syracuse and any other surrounding cities for a good OB. I’m also looking for a High Risk specialist or a lab who can do First Trimester screening for me. Please help!!!” Questions of such implicitly describe the physician information needs in situations where choosing a new physician who fits the patients’ expectation is a top priority. Investigating questions of physician information needs can advance the development of the patient-centered approach to physician selection.

2 Research Methods

This study analyzed 400 questions asking for recommendations for doctors collected from Health category of Yahoo Answers. Fifty questions of each of the eight medical specialties were purposefully solicited because these medical specialties usually have high demand for physician referral or recommendation. The eight medical specialties are gynecology and obstetrics, oncology, orthopedics and rehabilitation, dentistry, aesthetic medicine, neurology and neurosurgery, otolaryngology, and Chinese medicine. A combination of keywords, including “recommend,” “recommending,” “doctor(s),” “physician(s),” “recommended doctor(s)/physician(s),” “recommend a doctor/physician,” and “recommendations for doctor(s)/physician(s)” “would you (please) recommend a doctor/physician” were used to search for requests asking for recommendations for doctors. These keywords were typed into the search box of Health category of Yahoo! Answers, and names of the eight medical specialties were also used as keywords to filter and select only questions that met the criteria. The keyword search process was performed repeatedly until fifty questions for each of the eight medical specialties were fully collected.

Each question requesting for recommendations for doctors was numbered and classified into the medical specialty it belonged to, and was reasoned and designated as patients’ and caregivers’ articulated physician information needs. An Excel spreadsheet for each medical specialty was created and fifty questions for each specialty were recorded into the spreadsheets. For the purpose of this study, a unit of observation was the question requesting for recommended doctors, and a unit of analysis was a text chunk represented a single descriptor that an asker used to describe his or her physician information needs. The questions were analyzed with procedures of Grounded Theory, including open coding, axial coding, and selective coding [12]. With this method, coding categories were derived directly and inductively from the raw data- questions of request for recommended physicians. A classification model of descriptions of physician information needs was established based on the results of qualitative Grounded Theory analysis. Then the classification model was used as the coding schemes for the second stage of data analysis- quantitative content analysis. Similar to the procedures developed by [13], each question was coded for features of information need descriptions, and certain features were further subdivided into additional attributes. Percentage distribution of each attribute was determined as the ratio of the total number of observed frequency divided by the total of 400 questions.

The principal investigator along with two coders then performed constant comparison of Grounded Theory on recurring themes extracted from the questions. In the data analysis process, first, the researchers scanned through all the questions to gain familiarity with the requests. The characteristics found in the literature as well as the concepts derived from our own medical experiences were used as the initial coding schemes. Second, two coders read all the questions line-by-line, and unitized the concepts into themes, that is, when a new concept was mentioned outside of the original coding schemes, a new code was created and added to the existing coding template. For example, the idea of “language spoken” was not in the original coding scheme, it was later added to the coding schemes because it emerged from the data analysis process. Two coders independently coded 240 (60 %) questions, and they achieved high inter-coder reliability (kappa range 0.8–1.0 across codes); thus the principle investigator coded the rest of the questions. Then quantitative analysis was conducted with frequency count and percentage distribution to determine the patterns and characteristics of how patients conceptualized and articulated their physician information needs. In addition to analyzing the characteristics that are exhibited in 400 questions from eight medical specialties, this study also investigated the characteristics that are inherent across medical specialties.

3 Preliminary Findings

After analyzing 400 questions, the physician information needs are classified into four dimensions: “physician-related”, “patient-related”, “illness and disease-related”, and “institution and procedural-related”. Each question may at the same time address multiple types of attributes (Table 1).

Table 1. The attributes of four dimensions
Full size table

3.1 Physician-Related Dimension

Attributes of Physician-Related Dimension.

There are eight attributes in this dimension, which includes the medical specialty (95 %), skills and expertise (40 %), reputation (17 %), personality (15 %), practice (11 %), bedside manner (9 %), language spoken (1 %), and physicians’ gender (1 %) (Fig. 1). Most of the questions are about “physician’s medical specialty”, for instance, “My friend is diagnosed with cancer. Pls recommend good oncologist in Malaysia. Also let me have the contact No. & address”. The second highest is “physician skills and expertise”. For example, a patient addressed “I live in Texas and I am looking for a dentist who is excellent in porcelain veneers and is affordable”. People directly name the medical specialty of the physicians they consider would be a good fit.

Fig. 1.
figure 1

Attributes of physician-related dimension

Full size image

Attributes of Physician-Related Dimension by Medical Specialty.

Attributes in different medical specialties show diverse characteristics. “Physician’s medical specialty” is the most mentioned attributes in all medical specialties, which suggests that when people have need for physician for their health problems, they would describe medical specialty for the physician wanted for their health situations. For instance, “My friend diagnosed with cancer. Pls recommend good oncologist in Malaysia.” Second, “physician skills and expertise” is also a common question in most of medical specialties except Obstetrics and Gynecology. Third, “language spoken” and “physician gender” are less asked than the other attributes, but they both are shown in Aesthetic medicine and Obstetrics and Gynecology, which means that questions raised by people who have these kinds of problems might intend to ask for the physician’s language spoken ability or mention patient’s gender. For example, “my friend wants to come to Taiwan and has a cosmetic surgery, does anyone recommend doctors who speaks English well in Taipei?” or “know of a good female obstetrician in Mississauga?”

3.2 Patient-Related Dimension

Attributes of Patient-Related Dimension.

This dimension includes four attributes centered on patients’ gender (31 %), patients’ medical history (27 %), patients’ location (15 %), and patients’ age (5 %) (Fig. 2). Attributes that account for the highest proportion than the others are gender and medical history. Questions of gender are such as “I am 2- 3 weeks pregnant and I am new to this place…please share your experience with different doctors here.” On the other hand, questions of the medical history are, for example, “my little sister (age 32) has had brain tumors over the last 14 years with at least 12 surgeries being done for removal of the tumors, two of which cant be removed due to their location.”

Fig. 2.
figure 2

Attributes of patient-related dimension

Full size image

Attributes of Patient-Related Dimension by Medical Specialty.

In this dimension, “patient’s medical history” and “experience of treating patient of a particular gender” are the two most common attributes in all medical specialties. So it can be understood that “patient’s medical history” and “experience of treating patient of a particular gender” are usually discussed. For instance, “My little sister (age 32) has had brain tumors over the last 14 years with at least 12 surgeries being done for removal of the tumors , two of which can’t be removed due to their location.” is a typical question represented the attribute of “patient’s medical history”. And “my mom broke her bones, please recommend a good doctor near Wanhua.” is the example for attribute of “experience of treating patient of a particular gender.” On the other hand, medical specialty with higher instances of “experience of treating patient of a particular age” attribute, such as Dentistry can be classified as everyday special practice which means that the health problem would occur at patient of any age. So people may ask for a doctor who has such experience. For example, “could you recommend dentist for children in Tucheng? My child is about 1 year old. ” It’s also obvious that for patients requesting for recommendation of Obstetrics and Gynecology physician, they would need more location information, such as “I’m trying to have a baby and it just won’t work. Is there a doctor that you recommend? I live in New Hampshire. ”

3.3 Illness and Disease-Related Dimension

Attributes of Illness And Disease-Related Dimension.

There are two attributes in Illness and Disease-related dimension (Fig. 3); these are illness and disease (64 %), and symptoms (50 %). Patients usually name and specify the medical terms in the question. For example, as “I have been having horrible migraines and need to see a physician.” on attribute of illness and disease (64 %) or explain the situation he/she encounter, like “I spit up blood. Not cough, spit . And 2 days ago, I passed out ” on attributes of symptoms (50 %).

Fig. 3.
figure 3

Attributes of patient-related dimension

Full size image

Attributes of Illness and Disease-Related Dimension by Medical Specialty.

Attribute of “illness and disease” is used more than “symptoms” in most of medical specialties except Neurology and Neurosurgeon and Orthopaedics. People raised their question with exact illness terms often, such as “does anyone know authoritative doctor for Leukemia in Kaohsiung?” and the later one indicates that there are descriptions for illness in the questions, for example, “I’ve got some health issue recently, like dizzy, having head ache, and want to puke . Can anyone suggest a hospital in Taoyuan?”

3.4 Institution and Procedural-Related Dimension

Attributes of Institution And Procedural-Related Dimension.

This dimension contains seven attributes: medical treatment (72 %), practice location (63 %), medical cost of caring (9 %), classification of healthcare facility (8 %), practice hours (6 %), with particular medical equipment and supplies (2 %), and insurance coverage (2 %). The attribute account for highest proportion is medical treatment (72 %). An example of questions of medical treatment is “I am looking to see a doctor about my anxiety. I would like to find a doctor who will prescribe me with Xanax .” The second most commonly found attribute is practice location, and an example could be “We are looking for a decent family doctor in the Fayetteville, NC area . Any one has any recommendations?” (Fig. 4).

Fig. 4.
figure 4

Attributes of institution and procedural-related dimension

Full size image

Attributes of Institution and Procedural-Related Dimension by Medical Specialty.

In all medical specialties, attributes of “medical treatment” and “practice location” are asked more frequently than the others. People meet problems in Aesthetic medicine, Dentistry and Oncology raised questions with higher proportion of attribute of medical treatment. However, attribute of “practice location” is higher than “medical treatment” in Chinese medicine and Otolaryngological. For example, “Will anyone recommend a considerate doctor who had done laparotomy surgery ?” represents the characteristic of attributes of medical treatment. So it can be explained that people who have questions for Aesthetic medicine, Dentistry and Oncology, may focus on the medical treatments. On the other hand, people will value the importance of attribute of practice location when they have problems in Chinese medicine and Otolaryngological, such as “I’ve got a cold, can anyone recommend a doctor in old town in Tainan ?”

4 Implications of the Study

The major significance of this study differs from past research on patients’ information needs of physician selection lies in the contextualized nature of the questions analyzed. This study focuses on uncovering the complexity, uncertainty, and ambiguity of physician information retrieval by analyzing the nature of physician information needs from the users’ perspective, and re-examining the alternative attributes of the information organization for physician referral and recommendation systems. The attributes of the articulated of physician information needs derived from data analysis can be considered as the criteria that a patient or caregiver may follow when trying to find a “good” doctor for his health needs. The criteria may inform practice in designing systems for context-based classification model for physician recommender service.