Abstract

Objective

To determine whether physicians who interact with their patients between office visits using secure messaging and phone provide better care for patients with diabetes when controlling for physician, patient and care center characteristics.

Design

Retrospective study.

Setting

Kaiser Permanente Mid-Atlantic States.

Participants

174 Primary Care Physicians.

Intervention

We modeled the relationship between communication via secure messaging or phone communication and Diabetes Recognition Program (DRP) scores with a Generalized Estimating Equations model. Covariates included physician age and panel size, patient age, race, income and number of comorbidities, and the population density around the care center.

Main Outcome Measure

DRP scores.

Results

Physicians whose patients were predominantly white or mixed race were more likely than other physicians to use secure messaging and phone with their patients between visits, but there was no significant association between such contacts and DRP scores (P> 0.1). In contrast, physicians with predominantly black or Hispanic patients had significantly higher DRP scores associated with the use of secure messaging (P< 0.01) and higher, though not statistically significant, DRP scores associated with the use of phone (P< 0.1). These associations were strongest for outcome measures such as HbA1c and lipid levels, and were weaker or nonexistent for process measures such as annual foot and eye exams.

Conclusions

The use of secure messaging, and, to a lesser extent, phone, appears to be associated with higher quality diabetes care, particularly among at-risk populations.

Introduction

Effective communication is a critical component of high-quality healthcare for chronic diseases and contributes to the delivery of evidence-based care. However, office visits offer limited communication opportunities due to the priority that frequently must be placed on acute problems and because of the high volume of care-guidelines that must be addressed [1, 2]. Due to the time limitations of office visits, the Institute of Medicine has listed healthcare communication outside office visits (both by phone and online) as a key rule for healthcare redesign as a way of supporting continuous patient–physician relationships [3] and increasing healthcare quality. An accumulating body of evidence supports this recommendation by demonstrating that patients who communicate with their healthcare team between face to face visits, either by phone or by secure messaging, have generally better healthcare outcomes (phone: [4–9]; online:[10, 11]).

Recent studies have suggested that physician communication behaviors may influence patients' willingness to communicate between office visits, particularly when it comes to online communication tools. Weppner [12] and Ralston [10] both found that patients were more likely to use patient–physician secure email communication (referred to as secure messaging) if their primary care physician (PCP) was a frequent user of secure messaging. Combined with the evidence suggesting that between-visit communication improves individual healthcare outcomes, these data suggest that physicians' who use, and encourage their patients to use, out-of-office communication tools may provide higher quality patient care.

In this study, we used a physician-level analysis to evaluate the association between between-visit patient–physician communication and a key measure of the delivery of high-quality healthcare for patients with diabetes: the Diabetes Recognition Program (DRP) scores. We focused on care for patients with diabetes as it requires frequent communication between physician and patient and is likely to be positively affected by factors that increase patient–physician interactions. We chose DRP scores as our measure of panel-level quality as the final score consists of a roll-up of a variety of outcome and process measures, allowing us to separately estimate the association between out-of-office communication methods and outcome vs. process measures.

Although the Institute of Medicine does not distinguish between phone and online communications in its recommendations, we also considered whether the use of phone vs. secure messaging might have a different impact on care quality. Phone contacts have been part of medical practice for many years to increase access and provide medical triage [4, 13–15]. Secure messaging is a newer mode of patient–physician communication that is expected to increase healthcare efficiency and improve communication between patients and physicians [16–19]. Since its introduction there has been strong consumer demand for online communication with physicians [20–22], and there is some evidence that patients may be willing to discuss issues online that they may be too embarrassed to bring up in person or over the phone [23]. However, traditional phone-based care continues to play a large role in patient–physician communication as many physicians prefer phone communication to secure messaging [24, 25], not all patients have access to the infrastructure required for online communication [26], and some patients continue to have security concerns [23, 24]. Thus, in addition to assessing the overall relationship between out-of-office communication and DRP scores, we also evaluated the relationships between DRP scores and the use of phone and secure messaging separately to determine if the individual communication methods had different effects.

Finally, because race has been shown to have an effect on diabetes care and outcomes [27–29], face-to-face office visit communication [30–32], and the use of online tools [33], we explicitly examined how the relationship between out-of-office communication and DRP scores depended on the racial profile of the patient panel. We separated our physician population into two groups: one group consisted of physicians' whose patients were predominantly black or Hispanic, while the other group was composed of physicians whose patients had a more racially mixed profile. Each of these groups was examined separately to determine if communication between office visits had the same effect in patient panels composed of different races/cultures.

Methods

Study design

This retrospective study compared the use of out-of-office communication (secure messaging and phone calls) with scores calculated for accreditation in the DRP while controlling for physician, patient and care center factors. The study was performed at Kaiser Permanente, Mid-Atlantic States (KPMAS), using data abstracted from the electronic health record. The study was approved by the KPMAS Institutional Review Board on 9 January 2009.

Setting

KPMAS is an integrated delivery system that both insures and provides healthcare for over 485 000 members at 30 locations. The 950 physicians at KPMAS are compensated on a salary basis and thus have no financial incentives to require office visits to deliver care if that care can safely and effectively be provided through mechanisms other than in-person visits, such as secure messaging and phone calls.

All physicians at KPMAS use an electronic health record, EpicCare Ambulatory EMR (Epic Systems, Madison, WI, USA) to manage and record all care. The EMR contains patient information regarding primary and specialist care, outpatient surgery, lab services, radiology and pharmacy, as well as data from phone and secure messaging interactions.

In addition, all patients have access to a secure password protected web-based personal health record (PHR) based on Epic System's MyChart system. The PHR provides patient–physician secure messaging, as well as patient access to laboratory and pharmacy information and online appointment scheduling.

Data for this study were obtained from three sources: a data warehouse (Clarity) that contains an extract of the EMR and PHR data, a separate data warehouse that indicates which prescriptions were picked up by patients, and claims data for care provided external to the delivery system.

Subjects

We evaluated 174 PCPs (97% Internal Medicine, 25% Family Practice and 13% Pediatrics; some PCPs have more than one primary care specialty leading to a total >100%). PCPs were included in the analysis if they were practicing with KPMAS during the entire period 1 January  2007 through 31 December 2008 and had at least 25 patients with diabetes on their patient panel during 2008 (the year of analysis), consistent with the DRP program requirements [34]. When measuring diabetes patient panel size, we included only patients who were registered with the PCP throughout the study time period.

We identified patients with diabetes through Kaiser Permanente's population management database, Permanente Online Interactive Network Tools (POINT). Inclusion criteria for diabetes patient identification in POINT, based entirely on outpatient data, were: Exclusion criteria were: From the initial pool identified through POINT, patients were selected for inclusion in the analysis if they were (1) continuous KPMAS members for a full year prior to their last PCP encounter in 2008, (2) diagnosed with diabetes prior to 1 January 2008 and (3) did not switch PCPs during the year prior to their last 2008 PCP encounter.

• Two or more outpatient ICD9 code of 250.xx (DIABETES MELLITUS; ICD9 codes validated through random audits of patient records and associated claims; Outpatient codes from ER, Obstetrics/Gynecology, Podiatry and Ophthalmology are excluded) OR.

• Hemoglobin A1c>7.5% or Fructosamine >319 OR.

• A dispensing record for an oral hypoglycemic (excludes Metformin, Exenatide, Pioglitazone, Rosiglitazone or Repaglanide only) or insulin OR.

• Manually added to the diabetes patient registry by a physician.

• AND Member age >18 years.

• Female member coded for polycystic ovary syndrome and identified by Metformin, Pioglitazone or Rosiglitazone only OR.

• Female member with a positive pregnancy test OR code for gestational diabetes 648.XX is temporarily excluded from the diabetes population for 18 months after the positive pregnancy test date or date of code OR.

• Member with oral steroid Rx OR coded for pre-diabetes (790.22) or impaired glucose tolerance (790.29) in the last 12 months and no ICD9 code for diabetes OR.

• Member identified by a prescription for Metformin, Exenatide, Pioglitazone, Rosiglitazone or Repaglanide only OR.

• Manually removed from the diabetes patient registry by a physician.

Measurement of outcome variable

We used DRP scores to estimate how well individual doctors care for their entire practice of patients with diabetes. The National Committee for Quality Assurance DRP program assesses performance on 10 clinical indicators including both outcome and process metrics, as listed in Appendix 1 [34], which are combined to produce one overall DRP score that ranges from 0 to 100.

To calculate DRP scores, health-related data were extracted from the EMR for all patients with diabetes that met the inclusion criteria on each PCP's patient panel. The time period for analysis was 1 January 2008 to 31 December 2008. For each selected patient, we noted the value of the most recent HbA1C, LDL and BP recorded within that time frame and calculated the proportion of values above or below the cut-off values for each physician. Occurrence of a foot examination, eye examination, nephropathy assessment and smoking assessment/counseling were binary variables; we evaluated the proportion of patients on each physician's panel with each of these health events in their medical records to calculate scores on these metrics. Scores on the individual metrics were summed. Figure 1 shows the distribution of DRP scores; scores ranged from 15 to 100, with a median value of 75.

Figure 1

Distribution of Diabetes Physician Recognition Program (DRP) scores across physicians.

Figure 1

Distribution of Diabetes Physician Recognition Program (DRP) scores across physicians.

Measurement of secure messaging and phone interactions

To evaluate how DRP scores are related to out-of-office care, we focused on two forms of out-of-office communication: the use of secure messaging and the use of phone calls for interactions between patients and PCPs.

Secure messaging

To estimate the use of secure messaging within a given PCP's diabetes patient panel, we measured the proportion of the PCP's diabetes patients that had one or more secure message interactions with the physician, regardless of whether the secure message was initiated by the patient or the provider. We chose not to restrict the analysis to physician-initiated secure messages as KPMAS physicians frequently request that their patients send them information by secure message between visits. Thus, a secure message initiated by a patient could have been prompted by the physician's request during an office visit, making it difficult to determine who truly initiated the interaction.

For descriptive purposes, we also measure what proportion of the patient panel has registered for kp.org, Kaiser Permanente's PHR system. Registration for kp.org is a necessary requirement before patients can exchange secure messages with their physicians.

Phone

To measure the frequency of phone interactions between physicians and their diabetes patients, we tracked the number of phone encounters in the EMR linked to both the patient and the PCP. We then calculated both the proportion of each diabetes patient panel that interacted with the physician by phone at least once during the time period of interest.

For both secure message and phone communication, interactions were counted if the visit provider id listed for the encounter was the same as the PCP's provider ID; however, many of the interactions may have been completed by nurses or clinic assistants working with the PCP (the ‘healthcare team’). Only secure messages and phone calls directed to the healthcare team (rather than administrative or urgent care staff) were included in the analysis. No distinction was made between secure messages that contained clinically oriented interactions and those that contained more general interactions such as requests for refills. However, patients could schedule appointments and request renewals of medications electronically using specific portal functions without sending secure messages. These types of interactions were not counted as messages.

Total out-of-office communications

We also evaluated the use of any communications outside of office visits (secure messaging or phone) by measuring the proportion of the physicians' diabetes patient panels that used any combination of phone and secure messaging to interact with their physicians.

Covariates of interest

Because physicians were not randomly assigned to a communication method, additional unobserved factors may have influenced both the use of out of office communication methods and the physicians' DRP scores. We considered three types of factors that might influence both physician–patient communication methods and physician DRP scores: care center characteristics, physician characteristics and patient characteristics.

Care center characteristics

Prior research has shown that patients in rural areas may have poorer health outcomes than patients in urban areas [35]. In addition, patients in rural areas who have to travel further to see their doctor face to face may be more likely to use out-of-office communication methods with their physicians. Thus, the locale of the care center may impact the observed relationship between communication methods and physician scores. We used the population density in the care center's zip code as a proxy for determining whether the physician practiced in an urban, suburban or rural environment.

Physician characteristics

Based on previous examinations of secure messaging adoption, younger physicians may be more willing to use less traditional methods of communication such as secure messaging [36]. A preliminary correlation analysis also indicated a relationship between physician age and DRP scores. Thus, we included physician age as a covariate in the analysis of the relationship between communications methods and DRP scores.

In addition, the number of patients physicians care for may impact their availability for out of office communications [36], as well as the level of care they are able to provide, leading us to include total panel size as a covariate.

Patient characteristics

Although physicians are the primary unit of analysis in this paper, DRP scores ultimately measure the health and health behaviors of the physicians' patients. While a physician can influence these characteristics, patient outcomes and care processes also depend largely on characteristics of the patients themselves, such as their overall health. Accordingly, we measured the average number of comorbid conditions for the patient panel, where comorbid conditions consisted of the following diagnoses registered in the POINT population management database: asthma, coronary artery disease, cardiovascular disease, chronic kidney disease, hypertension and osteoporosis. In addition, we calculated the average number of in-person PCP doctor's office visits for the diabetes patients on each physician panel.

In addition, considerable research has shown that patient demographic characteristics may be strongly related to both diabetes outcomes [29] and patient–physician communication [10, 12, 30, 37]. Thus, if there are significant differences in the demographics between the physicians' panels, those differences may influence the observed correlation between out-of-office communication and DRP scores. We evaluated three major demographic characteristics of each patient panel: average age, racial profile and economic profile. Age was extracted from the Clarity data warehouse. Race and socioeconomic status were not well-captured in the Clarity data warehouse. To estimate these characteristics, we geocoded the patients' addresses to the census block group level and obtained the associated racial distribution and median household income from the ESRI Updated Census Demographics dataset for 2009. To obtain the overall racial profile of each physician's panel, we averaged over patients in each racial bin (White, Black, Asian-Pacific Islander, American Indian-Alaska Native, Hispanic and multiracial). We collapsed the overall distribution to a single measure by adding the proportion of patients who were Black and Hispanic. To measure the economic profile of the physicians' panels, we measured the average of the median household income across a given physician's patients.

A preliminary analysis of race determined that there were two subpopulations in the data, as shown in Fig. 2a. The first subgroup was composed of patient panels that were either predominantly white or racially mixed (%Black/Hispanic on patient panel <0.65), while the second subgroup included patient panels that were predominantly black and Hispanic (%Black/Hispanic on patient panel ≥0.65). Graphing DRP scores separately for the two subgroups revealed that predominantly white or mixed race panels tended to have higher DRP scores (Fig. 2b, top) than predominantly black and Hispanic panels (Fig. 2b, bottom). Because of the strong evidence for separate subpopulations in our data with different outcomes, we examined the relationship between communication methods and DRP scores separately for the two groups.

Figure 2

(a) Distribution of racial profile across physician panels. (b) Distribution of DRP scores for physicians with predominantly white or mixed race panels (top) and physicians with predominantly black/Hispanic panels (bottom).

Figure 2

(a) Distribution of racial profile across physician panels. (b) Distribution of DRP scores for physicians with predominantly white or mixed race panels (top) and physicians with predominantly black/Hispanic panels (bottom).

Statistical analyses

We used descriptive statistics to examine the characteristics of physicians and their patients. We then used linear regression models to explore the relationship between the use of secure messaging and phone and DRP scores while controlling for six physician, patient and care center covariates identified prior to analyses. Because all analyses were run separately for the two subgroups of physicians whose panels had different racial profiles, race was not included as a covariate in the regression model.

Regression models were estimated using generalized estimating equations (GEE model) to account for the clustering of physicians (and their patients) by care center. GEE models used an exchangeable working correlation matrix and robust covariance estimation. We ran three separate models investigating the effects of:

• the use of either type of out of office communication (secure messaging or phone calls);

• the use of secure messaging;

• the use of phone calls.

Results

A preliminary analysis comparing the use of out-of-office communication methods to DRP scores found that the use of both secure messaging and phone calls was significantly associated with improved DRP scores, as shown in Fig. 3 (Pearson's r correlation: secure messaging: r= 0.52, P< 0.01; phone: r= 0.22, P< 0.01). The goal of this study was to determine whether this association remains when confounding factors related to physician and patients are taken into account.

Figure 3

Relationships between the use of (a) secure messaging and (b) phone interactions and DRP scores. Each point in the plots represents one doctor's patient panel. The use of secure messaging or phone calls is represented by a value between 0 and 1 which indicates what proportion of the patient panel used that form of communication to interact with their physician. Dashed lines represent the regression line and the 95% confidence interval.

Figure 3

Relationships between the use of (a) secure messaging and (b) phone interactions and DRP scores. Each point in the plots represents one doctor's patient panel. The use of secure messaging or phone calls is represented by a value between 0 and 1 which indicates what proportion of the patient panel used that form of communication to interact with their physician. Dashed lines represent the regression line and the 95% confidence interval.

The analysis included a total of 174 PCPs that were separated into two subgroups based on the racial profile of their patient panels. One hundred and sixteen PCPs had patient panels that were predominantly white or racially mixed, while 58 PCPs had patients that were predominantly black or Hispanic. Table 1 compares the demographic characteristics of the two physician subgroups whose panels had different racial profiles. Compared with physicians who had predominantly white or mixed race panels, physicians with predominantly black or Hispanic patient panels tended to work in more urban areas and have patients with slightly more comorbidities and considerably lower incomes. The two physician groups were similar in age, number of total patients on their panels and average age of their patients.

Table 1

Characteristics of physician panels

Characteristic Panels that are predominantly white or mixed race Panels that are predominantly black or Hispanic
Mean (SD) Mean (SD)
Population densitya 4116 (1706) 6020 (3882)
Physician Age 47.8 (8.1) 46.6 (8.7)
Number of patients on panel 1309 (368) 1262 (374)
Number of patients with diabetes on panela 152 (63) 108 (44)
Average number of comorbiditiesa 2.2 (0.17) 2.3 (0.14)
Average age of panel 57.8 (2.3) 57.9 (2.5)
Median household income averaged across panela $90 359 ($18 014) $60 499 ($14 024)
Characteristic Panels that are predominantly white or mixed race Panels that are predominantly black or Hispanic
Mean (SD) Mean (SD)
Population densitya 4116 (1706) 6020 (3882)
Physician Age 47.8 (8.1) 46.6 (8.7)
Number of patients on panel 1309 (368) 1262 (374)
Number of patients with diabetes on panela 152 (63) 108 (44)
Average number of comorbiditiesa 2.2 (0.17) 2.3 (0.14)
Average age of panel 57.8 (2.3) 57.9 (2.5)
Median household income averaged across panela $90 359 ($18 014) $60 499 ($14 024)

aIndicates differences between the two groups of physician panels that were significant at the 0.01 level.

Table 2 shows the communication and online behaviors of the physicians and their patients for both subgroups of physicians. On average, physicians whose patient panels were predominantly white or mixed race were more likely to communicate with their patients between office visits using both phone (P< 0.01) and secure messaging (P< 0.01). These physicians also tended to have high DRP scores (80 ± 12, 1 SD). In this subgroup, a model evaluating whether the use of either out-of-office communication method (phone or secure messaging) was associated with improved DRP scores revealed no effect. Further analyses evaluating the separate effects of the use of secure messaging and phone also found no association with DRP scores. We concluded that out-of-office communication had no significant effects in this group, possibly due to consistently high DRP scores.

Table 2

Communication characteristics of the physicians and their patients.

Panels that are predominantly white or mixed race (%) Panels that are predominantly black or Hispanic (%)
Mean (SD) Mean (SD)
Proportion of patients that used either phone or secure messaginga 81 (11) 69 (16)
Proportion of patients that were members of kp.orga 49 (9) 30 (9)
Proportion of patients that used secure messaginga 33 (12) 18 (9)
Proportion of patients that used phonea 71 (12) 62 (17)
Panels that are predominantly white or mixed race (%) Panels that are predominantly black or Hispanic (%)
Mean (SD) Mean (SD)
Proportion of patients that used either phone or secure messaginga 81 (11) 69 (16)
Proportion of patients that were members of kp.orga 49 (9) 30 (9)
Proportion of patients that used secure messaginga 33 (12) 18 (9)
Proportion of patients that used phonea 71 (12) 62 (17)

aIndicates differences between the two groups of physician panels that were significant at the 0.01 level.

The remaining 58 PCPs had patient panels that were predominantly black or Hispanic. Although the two groups had approximately the same number of patients on their total patient panel, physicians whose patients were predominantly black or Hispanic cared for significantly more patients with diabetes (152 vs. 108, P< 0.01), and had lower DRP scores (51 ± 19, 1 SD). In this subgroup, after adjusting for correlated outcome data and controlling for six covariates associated with physician and panel characteristics, the GEE model revealed a significant, positive relationship between the use of out-of-office communication methods and DRP scores (P< 0.01). On average, a 0.1 increase in the proportion of the patient panel that shared out-of-office communication (including both phone and secure messaging) with their PCP was associated with a 1.6 unit increase in DRP scores.

Using separate GEE models for each mode of communication, we evaluated whether the use of secure messaging and the use of phone communications were associated with similar increases in DRP scores. Figure 4a shows the magnitude of the effects of mode of communication. After adjusting for clustering of data by care center and controlling for physician and panel characteristics, a 0.1 increase in the proportion of the patient panel that used secure messaging was associated with an increase in DRP scores of 4.7 (P< 0.01). In contrast, a 0.1 increase in the proportion of the patient panel that used phone to communicate with their PCP was only associated with an increase in DRP scores of 1.3 (P< 0.1).

Figure 4

Effects of out-of-office communication on DRP scores. Error bars indicate ±2 SE. (a) Strength of the association between communication method and DRP scores. (b) Comparison of the effects of secure messaging and phone on process metrics (light gray) and outcome metrics (dark gray).

Figure 4

Effects of out-of-office communication on DRP scores. Error bars indicate ±2 SE. (a) Strength of the association between communication method and DRP scores. (b) Comparison of the effects of secure messaging and phone on process metrics (light gray) and outcome metrics (dark gray).

We next evaluated whether the use of secure messaging and phone had larger effects on process or outcome measures. Figure 4b shows the changes in component scores associated with both the use of secure messaging and phone. We found that the use of secure messaging was strongly associated with improvements in outcome scores (P< 0.01). While the use of phone was also associated with improved outcomes scores, the relationship did not reach statistical significance (P< 0.1). In contrast, while the use of secure messaging was associated with an improvement in process scores (P< 0.01), the effect was relatively small, and the use of phone was not associated with any improvements in process scores.

Discussion

In this study of how communication outside office visits influences the quality of diabetes care, we found that the influence of out-of-office communications is strongly dependent on the profile of the physician's patient panel. For physicians whose patients were predominantly black or Hispanic, the use of secure messaging and phone calls between PCP physicians and patients with diabetes was associated with significantly better clinical quality of care measures. In contrast, for physicians whose panels were predominantly white, or which were racially mixed there was no clear association between the use of out-of-office communications and DRP scores.

In addition to the differences in the racial profile between the two subpopulations, there were several other significant differences. Physicians with racially mixed or predominantly white panels tended to have fewer patients with diabetes and were more likely to interact with them through both secure messaging and phone. Physicians in this group also tended to have high DRP scores, with relatively little variation among physicians. The combination of low variability in DRP scores and relatively high out-of-office communication penetration may have prevented us from finding a relationship in this group, even if the use of secure messaging and phone is useful to this subpopulation.

In contrast, physicians with predominantly black or Hispanic patients had considerably more variability in their DRP scores. In general, their patients were less likely to have registered to use the patient portal and use secure messaging. However, when physicians in this group did use secure messaging with their patients, it tended to be associated with higher DRP scores, even after controlling for panel size and patient demographics such as age and income. This association suggests that, in this subgroup, the use of secure messaging to communicate between office visits may contribute to the delivery of high-quality diabetes care.

The effect of physician–patient phone contact in this subpopulation was less clear. Although the use of phone was associated with improved DRP scores, the effect was relatively small and did not reach significance. The relatively small effect may point to the frequent use of phone for contacts that are less medically meaningful, such as checking the status of an appointment. While our counts of phone use excluded calls made directly to or by administrative staff, some patients may contact their doctor directly for administrative matters. These contacts are less likely to occur by secure messaging, since people using kp.org for secure messaging also have online access to administrative information about appointments.

Within the physician group that had predominantly black or Hispanic patients, non-office-based interactions were primarily associated with changes in outcome measures such as HbA1C values, cholesterol levels and blood pressure control. The use of secure messaging and phone calls had a relatively weak (secure messaging) or no (phone) association with process measures such as the presence of an eye or foot examination, nephropathy assessment or smoking cessation counseling. Although we did not explore the reasons for this difference, we note that most of these process measures require in-person visits, and are usually performed as secondary services at office visits. For this reason, these measures cannot be fulfilled through non-office-based communications, whereas adjustments in medication and reminders for lab tests that affect the outcome measures are amenable to non-office-based interactions.

Based on evidence that the majority of secure messaging contacts were initiated by patients (data not shown), we believe the associations described in this paper may be due to a strong patient-driven effect. For example, because the asynchronous nature of secure messaging avoids problems such as ‘phone tag’, patients may be more likely to contact their doctor if this tool is available to them. However, the effect is unlikely to be solely patient-driven: the variability in the use of secure messaging across physicians is positively associated with DRP scores, a physician-based measure. This association suggests that physicians may be influencing their patients' use of secure messaging, possibly by actively encouraging their patients to use this communication tool to provide health-related information between face-to-face visits. Disentangling the influences of patients and physicians on the use of out-of-office communication tools such as secure messaging is an important avenue for future research.

Further research is needed to understand the differences between secure messaging and phone calls for patient–physician interaction. Several studies have suggested that the use of secure messaging may be able to reduce the number of office visits to a greater extent than phone contacts alone ([16, 19] but see [38]). This finding suggests that secure messaging and phone calls may be used differently in the healthcare process. In addition, research shows that patients do not like contacting their physicians by phone and physicians often find playing phone tag with their patients frustrating [24]; however, physicians and patients continue to interact by phone significantly more often than by secure messaging. Future studies of the relative use of the two communication methods by physicians and patients may provide insight into these issues, as would an analysis of the subjects of secure messages and phone calls.

There are two major limitations of this study. The first is imposed by the cross-sectional design, which limits the ability to interpret causality based on the current results. Physicians who used out-of-office communication tools more frequently might have some common underlying factor that leads to both increased physician–patient communication and improved outcomes. In particular, physicians with better patient care skills may be more likely to encourage their patients to contact them between visits, and the generally better patient care they provide may translate into better DRP scores across their patient panel. Future research will explore the relationship between physicians' patient care skills and the use of communication tools.

Another possible explanation for the relationship between secure messaging and physician care scores is that patients who use secure messaging are simply more adept at managing their diabetes. In this case, physicians with a higher proportion of patients willing to use secure messaging would tend to have higher DRP scores, independent of the quality of care the physicians provide. Future research evaluating DRP scores based only on patients with equal propensity to use secure messaging could help to address this question.

A second major limitation of this study is the lack of content analysis of the secure message and phone interactions. Phone and secure message communications may have different levels of clinical content; for example, phone messages may be more likely to contain administrative matters such as questions about appointment status. Further research will compare the content of these types of communications to determine whether they are used for similar purposes.

Conclusions

In some groups, the use of secure messaging and to a lesser extent, phone, is associated with improvements in the quality of care for diabetes, as measured by DRP scores. These associations suggest that a) non-office-based interactions can improve care for patients with diabetes, and b) a greater emphasis needs to be placed on increasing interactions with vulnerable populations. More research is needed to understand how different types of communication methods impact different populations, and whether different communication types are used for different purposes.

Funding

This research was supported by the Mid-Atlantic Permanente Medical Group, Rockville, MD, USA. These results have been presented in part at the JAMIA Spring Congress, 2009.

Acknowledgements

We thank Michael Rowe and Jaewon Ryu for comments on the manuscript, and Adil Raqibuddin for his assistance with developing the data extraction programs for the Diabetes Recognition Program scores.

References

1
Østbye
T
Yarnall
KSH
Krause
KM
, et al.  .
Is there time for management of patients with chronic diseases in primary care?
Ann Fam Med
,
2005
, vol.
3
(pg.
209
-
14
)
2
Yarnall
KSH
Pollak
KI
Østbye
T
, et al.  .
Primary care: is there enough time for prevention?
Am J Public Health
,
2003
, vol.
93
(pg.
635
-
41
)
3
Institute of Medicine
Crossing the quality chasm
2001
4
Car
J
Sheikh
A
Telephone consultations
Br Med J
,
2003
, vol.
326
(pg.
966
-
9
)
5
Pinnock
H
L
S
, et al.  .
Accessibility, clinical effectiveness, and practice costs of providing a telephone option for routine asthma reviews: phase IV controlled implementation study
Br J Gen Pract
,
2007
, vol.
57
(pg.
714
-
22
)
6
Piette
JD
Weinberger
M
McPhee
SJ
The effect of automated calls with telephone nurse follow-up on patient-centered outcomes of diabetes care: a randomized, controlled trial
Med Care
,
2000
, vol.
38
(pg.
218
-
30
)
7
Wasson
J
Gaudette
C
Whaley
F
, et al.  .
Telephone care as a substitute for routine clinic follow-up
J Am Med Assoc
,
1992
, vol.
267
(pg.
1788
-
93
)
8
Gruffydd-Jones
K
Hollinghurst
S
Ward
S
, et al.  .
Targeted routine asthma care in general practice using telephone triage
Br J Gen Pract
,
2005
, vol.
55
(pg.
918
-
23
)
9
Courtney
M
Edwards
H
Chang
A
, et al.  .
Fewer emergency readmissions and better quality of life for older adults at risk of hospital readmission: a randomized controlled trial to determine the effectiveness of a 24-week exercise and telephone follow-up program
J Am Geriatr Soc
,
2009
, vol.
57
(pg.
395
-
402
)
10
Ralston
JD
Rutter
CM
Carrell
D
, et al.  .
Patient use of secure electronic messaging within a shared medical record: a cross-sectional study
J Gen Intern Med
,
2009
, vol.
24
(pg.
349
-
55
)
11
Zhou
YY
Kanter
MH
Wang
JJ
, et al.  .
Improved quality at Kaiser Permanente through e-mail between physicians and patients
Health Aff (Millwood)
,
2010
, vol.
29
(pg.
1370
-
5
)
12
Weppner
WG
Ralston
JD
Koepsell
TD
, et al.  .
Use of a shared medical record with secure messaging by older patients with diabetes
Diabetes Care

[Internet]. 2010 Aug 25; Available from http://www.ncbi.nlm.nih.gov/pubmed/20739686 (26 October 2010, date last accessed).
13
Bunn
F
Byrne
G
Kendall
S
The effects of telephone consultation and triage on healthcare use and patient satisfaction: a systematic review
Br J Gen Pract
,
2005
, vol.
55
(pg.
956
-
61
)
14
Delichatsios
H
Callahan
M
Charlson
M
Outcomes of telephone medical care
J Gen Intern Med
,
1998
, vol.
13
(pg.
579
-
85
)
15
McKinstry
B
Watson
P
Pinnock
H
, et al.  .
Telephone consulting in primary care: a triangulated qualitative study of patients and providers
Br J Gen Pract
,
2009
, vol.
59
(pg.
e209
-
218
)
16
Zhou
YY
Garrido
T
Chin
HL
, et al.  .
Patient access to an electronic health record with secure messaging: impact on primary care utilization
Am J Manag Care
,
2007
, vol.
13
(pg.
418
-
24
)
17
Dixon
RF
Enhancing primary care through online communication
Health Aff (Millwood)
,
2010
, vol.
29
(pg.
1364
-
9
)
18
Wallwiener
M
Wallwiener
CW
Kansy
JK
, et al.  .
Impact of electronic messaging on the patient-physician interaction
J Telemed Telecare
,
2009
, vol.
15
(pg.
243
-
50
)
19
Bergmo
TS
Kummervold
PE
Gammon
D
, et al.  .
Electronic patient-provider communication: will it offset office visits and telephone consultations in primary care?
Int J Med Inform
,
2005
, vol.
74
(pg.
705
-
10
)
20
Singh
H
Fox
SA
Petersen
NJ
, et al.  .
Older patients'enthusiasm to use electronic mail to communicate with their physicians: cross-sectional survey
J. Med. Internet Res
,
2009
, vol.
11
pg.
e18

21
Harris Interactive
Patient/physician online communication: many patients want it, would pay for it, and it would influence their choice of doctors and health plans
Health Care News
,
2002
, vol.
2
(pg.
1
-
4
)
22
D
Kujala
J
Ilvonen
K
The impact of patient-physician web messaging on healthcare service provision
Int J Med Inform
,
2008
, vol.
77
(pg.
17
-
23
)
23
Houston
TK
Sands
DZ
Jenckes
MW
, et al.  .
Experiences of patients who were early adopters of electronic communication with their physician: satisfaction, benefits, and concerns
Am J Manag Care
,
2004
, vol.
10
(pg.
601
-
8
)
24
Hassol
A
Walker
JM
Kidder
D
, et al.  .
J Am Med Inform Assoc
,
2004
, vol.
11
(pg.
505
-
13
)
25
Boukus
ER
Grossman
JM
O'Malley
AS
Physicians slow to e-mail routinely with patients
Issue Brief Cent Stud Health Syst Change
,
2010
, vol.
134
(pg.
1
-
5
)
26
Beckjord
EB
Finney Rutten
LJ
Squiers
L
, et al.  .
Use of the internet to communicate with health care providers in the United States: estimates from the 2003 and 2005 Health Information National Trends Surveys (HINTS)
J Med Internet Res
,
2007
, vol.
9
pg.
e20

27
Kirk
JK
D'Agostino
RB
Bell
RA
, et al.  .
Disparities in HbA1c levels between African-American and non-Hispanic white adults with diabetes: a meta-analysis
Diabetes Care
,
2006
, vol.
29
(pg.
2130
-
6
)
28
Harris
MI
Klein
R
Cowie
CC
, et al.  .
Is the risk of diabetic retinopathy greater in non-Hispanic blacks and Mexican Americans than in non-Hispanic whites with type 2 diabetes? A U.S. population study
Diabetes Care
,
1998
, vol.
21
(pg.
1230
-
5
)
29
Heisler
M
Smith
DM
Hayward
RA
, et al.  .
Racial disparities in diabetes care processes, outcomes, and treatment intensity
Med Care
,
2003
, vol.
41
(pg.
1221
-
32
)
30
Schouten
BC
Meeuwesen
L
Cultural differences in medical communication: a review of the literature
Patient Educ Couns
,
2006
, vol.
64
(pg.
21
-
34
)
31
Beach
MC
Saha
S
Korthuis
PT
, et al.  .
Patient-Provider Communication Differs for Black Compared to White HIV-Infected Patients
AIDS Behav

[Internet]. 2010 Jan 12; Available from http://www.ncbi.nlm.nih.gov/pubmed/20066486 (28 October 2010, date last accessed)
32
Diette
GB
Rand
C
The contributing role of health-care communication to health disparities for minority patients with asthma
Chest
,
2007
, vol.
132

5 Suppl
(pg.
802S
-
9S
)
33
Lorence
DP
Park
H
Fox
S
Racial disparities in health information access: resilience of the Digital Divide
J Med Syst
,
2006
, vol.
30
(pg.
241
-
9
)
34
National Committee for Quality Assurance
Diabetes Care
,
2008

35
Hale
NL
Bennett
KJ
Probst
JC
Diabetes care and outcomes: disparities across rural America
J Community Health
,
2010
, vol.
35
(pg.
365
-
74
)
36
Brooks
RG
Menachemi
N
Physicians'use of email with patients: factors influencing electronic communication and adherence to best practices
J Med Internet Res
,
2006
, vol.
8
pg.
e2

37
Street
RL
Gordon
HS
Ward
MM
, et al.  .
Patient participation in medical consultations: why some patients are more involved than others
Med Care
,
2005
, vol.
43
(pg.
960
-
9
)
38
Harris
LT
Haneuse
SJ
Martin
DP
, et al.  .
Diabetes quality of care and outpatient utilization associated with electronic patient-provider messaging: a cross-sectional analysis
Diabetes Care
,
2009
, vol.
32
(pg.
1182
-
7
)

Appendix 1

Description of metrics used to calculate the DRP score.

Type of metric Diabetes recognition measure Required % of patients on panel Weight
Outcome HbA1C Control >9.0 ≤15 15
Outcome HbA1C Control <7.0 ≥40 10
Outcome Blood Pressure Control ≥140/90 mm Hg ≤35 15
Outcome Blood Pressure Control <130/80 mm Hg ≥25 10
Outcome LDL Control ≥130 mg/dl ≤37 10
Outcome LDL Control <100 mg/dl ≥36 10
Process Annual Eye Examination ≥60 10
Process Annual Foot Examination ≥80
Process Annual Nephropathy Assessment ≥80
Process Smoking Status & Cessation Advice or Treatment ≥80 10
Total points 100
Type of metric Diabetes recognition measure Required % of patients on panel Weight
Outcome HbA1C Control >9.0 ≤15 15
Outcome HbA1C Control <7.0 ≥40 10
Outcome Blood Pressure Control ≥140/90 mm Hg ≤35 15
Outcome Blood Pressure Control <130/80 mm Hg ≥25 10
Outcome LDL Control ≥130 mg/dl ≤37 10
Outcome LDL Control <100 mg/dl ≥36 10
Process Annual Eye Examination ≥60 10
Process Annual Foot Examination ≥80
Process Annual Nephropathy Assessment ≥80
Process Smoking Status & Cessation Advice or Treatment ≥80 10
Total points 100