The Technical Performance and Clinical Feasibility of Telecolposcopy

The goal of colposcopy is to identify women with high-grade disease through appropriately targeted biopsy and to correlate the cytologic-colposcopic-histologic results into a cohesive and appropriate management strategy. Approximately 80% of gynecologists and 10% of family physicians in New Hampshire currently provide access to colposcopy. These physicians have a desire to network to improve their colposcopic accuracy.¹ Little work has been published either to define adequate access to colposcopy for any population of women or to improve colposcopic accuracy. Active maintenance and improvement of colposcopic skills are important, because they deteriorate after initial training. Several studies have reported lower than 50% sensitivity for detecting the quadrant of the cervix where the cervical intraepithelial neoplasia (CIN) is present.^2-4 “See and treat” electrosurgical loop excision procedures have resulted in up to 50% of the specimens being histologically normal.^5-9 These high rates of inaccuracy happen because of colposcopists’ inability to distinguish normal from abnormal cervical tissue. Without a dedicated desire to maintain and improve their skills, colposcopists may inadvertently overtreat women on the pretext of preventing cervical cancer.

One potential method for improving colposcopic recognition skills is the development of a network of telecolposcopists who share patient histories, colposcopic images, and histology correlations on a routine basis. Before determining whether this network would improve colposcopic care, it must first be shown that telecolposcopic images can be transmitted to and recognized accurately at the reception site.

Radiologists, dermatologists, pathologists, and psychiatrists^10-13 have accurately relayed image transmissions. The fields of otorhinolaryngology, ophthalmology, cardiology, pulmonology, and emergency medicine have also reported telemedicine capabilities during the past few years. Teleobstetrics has demonstrated the feasibility of remote fetal ultrasonography and home monitoring of uterine activity accurately portraying fetal anomalies and uterine contractility.^14-17 Colposcopy provides similar visual information about a woman’s gynecologic health. The parameters for successful telemedical image interpretation have been established for these other fields by studying the correlation between the telemedicine image and the conventional diagnostic modality.^11-13,18-23

The purpose of our project was to demonstrate the technical performance and clinical feasibility of a low-end telecolposcopic system in a rural mountainous environment. Our primary goal was to determine if the transmission of colposcopic images was technically feasible and would be accurately received in an environment where the ground terrain prohibits radio wave and some satellite transmissions. Our secondary goals were to determine whether the telecolposcopy system could be incorporated into a busy medical office with minimal disruption and to determine how it affects the woman’s perceptions of the colposcopic examination. We included several questions to determine how far each woman was willing to travel to see the expert in person rather than participate in telecolposcopy.

Methods

The Review and Remote Site Selection

The review site was based at the Dartmouth Hitchcock Medical Center (DHMC), where an experienced colposcopist reviewed all images. Two remote sites were chosen for this feasibility study: The first site was a single-physician practice located in rural New Hampshire; the second site was an urban residency training clinic remote from DHMC.*

The telecolposcopic system was designed to maximize the quality of digitized images viewed during the examination and deliver those images electronically to the review center. The computer system was installed by the computer support personnel at each remote site in less than 2 hours, and included a brief introduction to the software and 3 to 5 trial runs with mock patients. Computer support personnel visited both sites one additional time during the study to answer software, hardware, and transmission questions. Our study received approval from the Committee on Protection of Human Subjects from Dartmouth Medical School.

A provider survey was designed to document the learning curve of the computer-based digital colposcopic system in an office with limited computer support available (site 1). A 23-question patient satisfaction survey was administered in 2 parts to assess the acceptability of telecolposcopy to women undergoing a colposcopic examination.

Implementation of the Telecolposcopic System

Any woman 18 years or older and scheduled for colposcopy was eligible for enrollment in the study between October 1997 and May 1998. There were no other inclusion or exclusion criteria. After consent was obtained, the woman was given patient education materials to read and asked to complete the first part of the survey. The colposcopic examination was completed as usual, except that images were focused and captured throughout the examination to document the visualization of the entire squamocolumnar junction (SCJ), transformation zone, and any visible lesions in the anogenital area. The colposcopist entered a follow-up plan in the computerized patient record. The images and history were then printed for the patient’s chart. The patient was asked to complete the second part of the satisfaction questionnaire before dismissal, and the physician was asked to complete the provider satisfaction questionnaire. The woman’s cytology and histology results were processed and read by the site’s routine pathology laboratory and were entered into the study database. The review site colposcopist was blinded to the primary colposcopist’s impressions, the woman’s cytology and histology, and any information on the questionnaires.

Statistical Analysis

Our study has a 90% power to determine a statistically significant k correlation, at the 0.05 2-tailed level, between the remote and review site colposcopists’ identification of the SCJ and their impressions of lesion severity and between each colposcopist’s impression and pathology. We used k correlation, independent and paired t tests, and descriptive statistics in this study. All statistics were calculated with Statistica (StatSoft, Tulsa, Oklahoma).²⁴

Results

A total of 79 women participated in our study. Sites 1 and 2 captured images of 39 and 40 women, respectively, representing 100% of the women scheduled for colposcopy at each site during the study. All women were white and ranged in age from 18 to 57 years, with a mean age of 29.5 years (standard deviation=11.5). Thirty-five percent of the women received their colposcopy at their usual site of gynecologic health care; the others were referred from outside offices where colposcopy was not available. The [Table] shows the descriptive data of the women participating in the project: 79% had a history of previous abnormal anogenital lesions; 65% presented with either atypical squamous cells of undetermined significance or low-grade squamous intraepithelial lesion (LSIL) cytology, and approximately 20% had a CIN 2/3 lesion histologically.

Technical Feasibility

The number of images captured for each patient varied from 3 to 20, and the details of the patient’s gynecologic history varied from the current visit data to an extensive past medical and gynecologic history. The average time for transmission of one patient’s images was 6 minutes.

Technical feasibility was measured by the percentage of agreement between the review and both remote sites for colposcopic impressions and histology correlates. A k value greater than 0.75 denotes excellent agreement, between 0.4 and 0.75 is good agreement, and below 0.4 is marginal agreement.25 For all examinations, the physician at the remote sites attempted to capture the SCJ for transmission. The SCJ was noted as “completely visualized” or “not visualized.” There was 86% agreement in complete visualization of the SCJ between the review and the remote sites (k=.533; 95% confidence interval [CI], 0.276-0.791; P=.019).

Both remote and review site physicians classified the colposcopic cervical impression as normal, LSIL, or high-grade-squamous intraepithelial lesion (HSIL). The interobserver agreement for colposcopic impressions was 86% (k=.684; 95% CI, 0.544-0.825; P <.001), and 66% for colposcopic impressions with exact histology (k=.416; 95% CI, 0.323-0.509; P <.001). If agreement between colposcopic impression and histology was relaxed to within 1 degree of severity, there was an 86% interobserver correlation (k=0.78; 95% CI, 0.697-0.861; P <.001). The distinction between normal metaplastic epithelium and low-grade changes accounted for all the discrepancies.

In our study, the implementation of a telecolposcopy system within the current office structure was easily accomplished, and most women accepted the telecolposcopic examination.* A total of 90.5% of the women surveyed travel 25 miles or less to their usual site of general health care; 88.1% traveled 25 miles or less for their colposcopy appointment. When the women were asked to choose whether they would prefer to have a colposcopy done locally and have the images electronically sent to another physician to be reviewed or drive farther than usual to the nearest experienced colposcopist, 95.2% of the women stated that they would have their colposcopy locally with electronic submission if the experienced colposcopist were more than 25 miles away. For the 12% of women who traveled more than 25 miles for their scheduled colposcopy, all stated they would have colposcopy locally if the experienced colposcopist was farther away than their current colposcopist.

Discussion

The telecolposcopy system captured the SCJ extremely well. Our study had an interobserver correlation of SCJ visualization similar to that seen for comparisons between different readers of the same cervigram image.^26,27 To compare between review and remote sites, the colposcopist at site 1 determined the impression while at the colposcope, and the colposcopist at the review center used the same patient’s digital image. The interobserver agreement for our colposcopic image versus digital image colposcopic impressions (86%) was higher than that when both colposcopic impressions were read from colpophotographs.²⁸ Our interobserver correlations between the colposcopic impression and histology agreed with others’ correlations for exact histologic agreement^29,30 (66%) references and was within 1 degree of disease severity (86%) reference number.³¹ The difference in interobserver agreements for the colposcopic impression and histology correlates reflects the difficulty in determining normal metaplasia from low-grade disease, a more difficult distinction than that between normal and high-grade or cancer disease.^8,28

Implementation of any organizational change within a medical office can often be difficult. It is more easily accepted when the organizational change is part of a clinical study.³² Both remote sites adopted the full telecolposcopic system (patient history data and image collection) after our study, because the teaching system, consultation reports, and record documentation were superior to the previous systems. Our study initially gave them the support and structure to use the telecolposcopic system. Implicit in adopting a new computerized system is the availability of someone to troubleshoot software and hardware problems. A remote telecolposcopic system requires that the main technical support comes from the review center, with onsite local technical help available, if necessary. In our study, technical help provided by the review center successfully kept the remote telecolposcopy systems functional with minimal effort.

More than two thirds of the women in our study were referred from their usual site of gynecologic health care to the colposcopist, and most traveled 25 miles or less to keep the appointment. Other studies have shown that women will drive more than 200 miles to participate in health care screening activities.³³ We have shown that when given the choice of a local provider with telecolposcopy access or driving more than 25 miles for an experienced examination, 95% of the women preferred to receive their care from a local colposcopist. This implies that for women in rural communities, local access to colposcopy with telecolposcopic links is preferred to traveling to a distant center of colposcopic excellence. Future efforts at increasing access to colposcopy for women should be directed toward local telecolposcopic connections with a regional center of excellence for image referral.

Conclusions

We have demonstrated that telecolposcopy is feasible, can be implemented in mountainous rural office sites, and is acceptable to the women undergoing the examination. To move from a demonstration project to a regional telecolposcopic network, telemedicine standards established for all telespecialties must be followed, including appropriate licensure and encryption standards.³⁴ Following these standards, the telecolposcopic network has the potential to offer increased access to and improved accuracy of colposcopic services. In addition, as an educational tool telecolposcopic images can be used for medical student, resident, and continuing medical education instruction, as well as direct patient education. Colposcopic training within an obstetrics and gynecology or family medicine residency training program occupies a small proportion of the curriculum. Telecolposcopy can be used to assist residents and other health care providers located at remote sites by providing access to experienced faculty and a large database of patients to review. The digital cervical images, if focused correctly, can provide more detail than the current hand-drawn documentation, which may lead to better cytologic, colposcopic, and histologic correlation.⁹

Colposcopists who as individuals perform a small number of colposcopies can gain knowledge from communicating between practices with the other single colposcopists through this network. They can all then anonymously compare their diagnoses with the actual biopsy result as a voluntary quality assurance and improvement program. This benchmarking can provide the groundwork for national colposcopic standards, as has been done in other countries^35,36 and in the United States for mammographic image and histology correlations.³⁷ The telecolposcopic system could focus on increasing access to colposcopy and on improving colposcopic pattern recognition skills so that high-grade disease would be targeted for treatment, and the overtreatment of low-grade disease would be avoided.

Acknowledgments

The New Hampshire Health Care Transition Fund Grant 98136, the Koop Institute, the Robert Wood Johnson Foundation, and the Division of Gynecologic Oncology, Dartmouth Medical School, have supported this work.

The goal of colposcopy is to identify women with high-grade disease through appropriately targeted biopsy and to correlate the cytologic-colposcopic-histologic results into a cohesive and appropriate management strategy. Approximately 80% of gynecologists and 10% of family physicians in New Hampshire currently provide access to colposcopy. These physicians have a desire to network to improve their colposcopic accuracy.¹ Little work has been published either to define adequate access to colposcopy for any population of women or to improve colposcopic accuracy. Active maintenance and improvement of colposcopic skills are important, because they deteriorate after initial training. Several studies have reported lower than 50% sensitivity for detecting the quadrant of the cervix where the cervical intraepithelial neoplasia (CIN) is present.^2-4 “See and treat” electrosurgical loop excision procedures have resulted in up to 50% of the specimens being histologically normal.^5-9 These high rates of inaccuracy happen because of colposcopists’ inability to distinguish normal from abnormal cervical tissue. Without a dedicated desire to maintain and improve their skills, colposcopists may inadvertently overtreat women on the pretext of preventing cervical cancer.

One potential method for improving colposcopic recognition skills is the development of a network of telecolposcopists who share patient histories, colposcopic images, and histology correlations on a routine basis. Before determining whether this network would improve colposcopic care, it must first be shown that telecolposcopic images can be transmitted to and recognized accurately at the reception site.

Radiologists, dermatologists, pathologists, and psychiatrists^10-13 have accurately relayed image transmissions. The fields of otorhinolaryngology, ophthalmology, cardiology, pulmonology, and emergency medicine have also reported telemedicine capabilities during the past few years. Teleobstetrics has demonstrated the feasibility of remote fetal ultrasonography and home monitoring of uterine activity accurately portraying fetal anomalies and uterine contractility.^14-17 Colposcopy provides similar visual information about a woman’s gynecologic health. The parameters for successful telemedical image interpretation have been established for these other fields by studying the correlation between the telemedicine image and the conventional diagnostic modality.^11-13,18-23

The purpose of our project was to demonstrate the technical performance and clinical feasibility of a low-end telecolposcopic system in a rural mountainous environment. Our primary goal was to determine if the transmission of colposcopic images was technically feasible and would be accurately received in an environment where the ground terrain prohibits radio wave and some satellite transmissions. Our secondary goals were to determine whether the telecolposcopy system could be incorporated into a busy medical office with minimal disruption and to determine how it affects the woman’s perceptions of the colposcopic examination. We included several questions to determine how far each woman was willing to travel to see the expert in person rather than participate in telecolposcopy.

Methods

The Review and Remote Site Selection

The review site was based at the Dartmouth Hitchcock Medical Center (DHMC), where an experienced colposcopist reviewed all images. Two remote sites were chosen for this feasibility study: The first site was a single-physician practice located in rural New Hampshire; the second site was an urban residency training clinic remote from DHMC.*

The telecolposcopic system was designed to maximize the quality of digitized images viewed during the examination and deliver those images electronically to the review center. The computer system was installed by the computer support personnel at each remote site in less than 2 hours, and included a brief introduction to the software and 3 to 5 trial runs with mock patients. Computer support personnel visited both sites one additional time during the study to answer software, hardware, and transmission questions. Our study received approval from the Committee on Protection of Human Subjects from Dartmouth Medical School.

A provider survey was designed to document the learning curve of the computer-based digital colposcopic system in an office with limited computer support available (site 1). A 23-question patient satisfaction survey was administered in 2 parts to assess the acceptability of telecolposcopy to women undergoing a colposcopic examination.

Implementation of the Telecolposcopic System

Any woman 18 years or older and scheduled for colposcopy was eligible for enrollment in the study between October 1997 and May 1998. There were no other inclusion or exclusion criteria. After consent was obtained, the woman was given patient education materials to read and asked to complete the first part of the survey. The colposcopic examination was completed as usual, except that images were focused and captured throughout the examination to document the visualization of the entire squamocolumnar junction (SCJ), transformation zone, and any visible lesions in the anogenital area. The colposcopist entered a follow-up plan in the computerized patient record. The images and history were then printed for the patient’s chart. The patient was asked to complete the second part of the satisfaction questionnaire before dismissal, and the physician was asked to complete the provider satisfaction questionnaire. The woman’s cytology and histology results were processed and read by the site’s routine pathology laboratory and were entered into the study database. The review site colposcopist was blinded to the primary colposcopist’s impressions, the woman’s cytology and histology, and any information on the questionnaires.

Statistical Analysis

Our study has a 90% power to determine a statistically significant k correlation, at the 0.05 2-tailed level, between the remote and review site colposcopists’ identification of the SCJ and their impressions of lesion severity and between each colposcopist’s impression and pathology. We used k correlation, independent and paired t tests, and descriptive statistics in this study. All statistics were calculated with Statistica (StatSoft, Tulsa, Oklahoma).²⁴

Results

A total of 79 women participated in our study. Sites 1 and 2 captured images of 39 and 40 women, respectively, representing 100% of the women scheduled for colposcopy at each site during the study. All women were white and ranged in age from 18 to 57 years, with a mean age of 29.5 years (standard deviation=11.5). Thirty-five percent of the women received their colposcopy at their usual site of gynecologic health care; the others were referred from outside offices where colposcopy was not available. The [Table] shows the descriptive data of the women participating in the project: 79% had a history of previous abnormal anogenital lesions; 65% presented with either atypical squamous cells of undetermined significance or low-grade squamous intraepithelial lesion (LSIL) cytology, and approximately 20% had a CIN 2/3 lesion histologically.

Technical Feasibility

The number of images captured for each patient varied from 3 to 20, and the details of the patient’s gynecologic history varied from the current visit data to an extensive past medical and gynecologic history. The average time for transmission of one patient’s images was 6 minutes.

Technical feasibility was measured by the percentage of agreement between the review and both remote sites for colposcopic impressions and histology correlates. A k value greater than 0.75 denotes excellent agreement, between 0.4 and 0.75 is good agreement, and below 0.4 is marginal agreement.25 For all examinations, the physician at the remote sites attempted to capture the SCJ for transmission. The SCJ was noted as “completely visualized” or “not visualized.” There was 86% agreement in complete visualization of the SCJ between the review and the remote sites (k=.533; 95% confidence interval [CI], 0.276-0.791; P=.019).

Both remote and review site physicians classified the colposcopic cervical impression as normal, LSIL, or high-grade-squamous intraepithelial lesion (HSIL). The interobserver agreement for colposcopic impressions was 86% (k=.684; 95% CI, 0.544-0.825; P <.001), and 66% for colposcopic impressions with exact histology (k=.416; 95% CI, 0.323-0.509; P <.001). If agreement between colposcopic impression and histology was relaxed to within 1 degree of severity, there was an 86% interobserver correlation (k=0.78; 95% CI, 0.697-0.861; P <.001). The distinction between normal metaplastic epithelium and low-grade changes accounted for all the discrepancies.

In our study, the implementation of a telecolposcopy system within the current office structure was easily accomplished, and most women accepted the telecolposcopic examination.* A total of 90.5% of the women surveyed travel 25 miles or less to their usual site of general health care; 88.1% traveled 25 miles or less for their colposcopy appointment. When the women were asked to choose whether they would prefer to have a colposcopy done locally and have the images electronically sent to another physician to be reviewed or drive farther than usual to the nearest experienced colposcopist, 95.2% of the women stated that they would have their colposcopy locally with electronic submission if the experienced colposcopist were more than 25 miles away. For the 12% of women who traveled more than 25 miles for their scheduled colposcopy, all stated they would have colposcopy locally if the experienced colposcopist was farther away than their current colposcopist.

Discussion

The telecolposcopy system captured the SCJ extremely well. Our study had an interobserver correlation of SCJ visualization similar to that seen for comparisons between different readers of the same cervigram image.^26,27 To compare between review and remote sites, the colposcopist at site 1 determined the impression while at the colposcope, and the colposcopist at the review center used the same patient’s digital image. The interobserver agreement for our colposcopic image versus digital image colposcopic impressions (86%) was higher than that when both colposcopic impressions were read from colpophotographs.²⁸ Our interobserver correlations between the colposcopic impression and histology agreed with others’ correlations for exact histologic agreement^29,30 (66%) references and was within 1 degree of disease severity (86%) reference number.³¹ The difference in interobserver agreements for the colposcopic impression and histology correlates reflects the difficulty in determining normal metaplasia from low-grade disease, a more difficult distinction than that between normal and high-grade or cancer disease.^8,28

Implementation of any organizational change within a medical office can often be difficult. It is more easily accepted when the organizational change is part of a clinical study.³² Both remote sites adopted the full telecolposcopic system (patient history data and image collection) after our study, because the teaching system, consultation reports, and record documentation were superior to the previous systems. Our study initially gave them the support and structure to use the telecolposcopic system. Implicit in adopting a new computerized system is the availability of someone to troubleshoot software and hardware problems. A remote telecolposcopic system requires that the main technical support comes from the review center, with onsite local technical help available, if necessary. In our study, technical help provided by the review center successfully kept the remote telecolposcopy systems functional with minimal effort.

More than two thirds of the women in our study were referred from their usual site of gynecologic health care to the colposcopist, and most traveled 25 miles or less to keep the appointment. Other studies have shown that women will drive more than 200 miles to participate in health care screening activities.³³ We have shown that when given the choice of a local provider with telecolposcopy access or driving more than 25 miles for an experienced examination, 95% of the women preferred to receive their care from a local colposcopist. This implies that for women in rural communities, local access to colposcopy with telecolposcopic links is preferred to traveling to a distant center of colposcopic excellence. Future efforts at increasing access to colposcopy for women should be directed toward local telecolposcopic connections with a regional center of excellence for image referral.

Conclusions

We have demonstrated that telecolposcopy is feasible, can be implemented in mountainous rural office sites, and is acceptable to the women undergoing the examination. To move from a demonstration project to a regional telecolposcopic network, telemedicine standards established for all telespecialties must be followed, including appropriate licensure and encryption standards.³⁴ Following these standards, the telecolposcopic network has the potential to offer increased access to and improved accuracy of colposcopic services. In addition, as an educational tool telecolposcopic images can be used for medical student, resident, and continuing medical education instruction, as well as direct patient education. Colposcopic training within an obstetrics and gynecology or family medicine residency training program occupies a small proportion of the curriculum. Telecolposcopy can be used to assist residents and other health care providers located at remote sites by providing access to experienced faculty and a large database of patients to review. The digital cervical images, if focused correctly, can provide more detail than the current hand-drawn documentation, which may lead to better cytologic, colposcopic, and histologic correlation.⁹

Colposcopists who as individuals perform a small number of colposcopies can gain knowledge from communicating between practices with the other single colposcopists through this network. They can all then anonymously compare their diagnoses with the actual biopsy result as a voluntary quality assurance and improvement program. This benchmarking can provide the groundwork for national colposcopic standards, as has been done in other countries^35,36 and in the United States for mammographic image and histology correlations.³⁷ The telecolposcopic system could focus on increasing access to colposcopy and on improving colposcopic pattern recognition skills so that high-grade disease would be targeted for treatment, and the overtreatment of low-grade disease would be avoided.

Acknowledgments

The New Hampshire Health Care Transition Fund Grant 98136, the Koop Institute, the Robert Wood Johnson Foundation, and the Division of Gynecologic Oncology, Dartmouth Medical School, have supported this work.

The goal of colposcopy is to identify women with high-grade disease through appropriately targeted biopsy and to correlate the cytologic-colposcopic-histologic results into a cohesive and appropriate management strategy. Approximately 80% of gynecologists and 10% of family physicians in New Hampshire currently provide access to colposcopy. These physicians have a desire to network to improve their colposcopic accuracy.¹ Little work has been published either to define adequate access to colposcopy for any population of women or to improve colposcopic accuracy. Active maintenance and improvement of colposcopic skills are important, because they deteriorate after initial training. Several studies have reported lower than 50% sensitivity for detecting the quadrant of the cervix where the cervical intraepithelial neoplasia (CIN) is present.^2-4 “See and treat” electrosurgical loop excision procedures have resulted in up to 50% of the specimens being histologically normal.^5-9 These high rates of inaccuracy happen because of colposcopists’ inability to distinguish normal from abnormal cervical tissue. Without a dedicated desire to maintain and improve their skills, colposcopists may inadvertently overtreat women on the pretext of preventing cervical cancer.

One potential method for improving colposcopic recognition skills is the development of a network of telecolposcopists who share patient histories, colposcopic images, and histology correlations on a routine basis. Before determining whether this network would improve colposcopic care, it must first be shown that telecolposcopic images can be transmitted to and recognized accurately at the reception site.

Radiologists, dermatologists, pathologists, and psychiatrists^10-13 have accurately relayed image transmissions. The fields of otorhinolaryngology, ophthalmology, cardiology, pulmonology, and emergency medicine have also reported telemedicine capabilities during the past few years. Teleobstetrics has demonstrated the feasibility of remote fetal ultrasonography and home monitoring of uterine activity accurately portraying fetal anomalies and uterine contractility.^14-17 Colposcopy provides similar visual information about a woman’s gynecologic health. The parameters for successful telemedical image interpretation have been established for these other fields by studying the correlation between the telemedicine image and the conventional diagnostic modality.^11-13,18-23

The purpose of our project was to demonstrate the technical performance and clinical feasibility of a low-end telecolposcopic system in a rural mountainous environment. Our primary goal was to determine if the transmission of colposcopic images was technically feasible and would be accurately received in an environment where the ground terrain prohibits radio wave and some satellite transmissions. Our secondary goals were to determine whether the telecolposcopy system could be incorporated into a busy medical office with minimal disruption and to determine how it affects the woman’s perceptions of the colposcopic examination. We included several questions to determine how far each woman was willing to travel to see the expert in person rather than participate in telecolposcopy.

Methods

The Review and Remote Site Selection

The review site was based at the Dartmouth Hitchcock Medical Center (DHMC), where an experienced colposcopist reviewed all images. Two remote sites were chosen for this feasibility study: The first site was a single-physician practice located in rural New Hampshire; the second site was an urban residency training clinic remote from DHMC.*

The telecolposcopic system was designed to maximize the quality of digitized images viewed during the examination and deliver those images electronically to the review center. The computer system was installed by the computer support personnel at each remote site in less than 2 hours, and included a brief introduction to the software and 3 to 5 trial runs with mock patients. Computer support personnel visited both sites one additional time during the study to answer software, hardware, and transmission questions. Our study received approval from the Committee on Protection of Human Subjects from Dartmouth Medical School.

A provider survey was designed to document the learning curve of the computer-based digital colposcopic system in an office with limited computer support available (site 1). A 23-question patient satisfaction survey was administered in 2 parts to assess the acceptability of telecolposcopy to women undergoing a colposcopic examination.

Implementation of the Telecolposcopic System

Any woman 18 years or older and scheduled for colposcopy was eligible for enrollment in the study between October 1997 and May 1998. There were no other inclusion or exclusion criteria. After consent was obtained, the woman was given patient education materials to read and asked to complete the first part of the survey. The colposcopic examination was completed as usual, except that images were focused and captured throughout the examination to document the visualization of the entire squamocolumnar junction (SCJ), transformation zone, and any visible lesions in the anogenital area. The colposcopist entered a follow-up plan in the computerized patient record. The images and history were then printed for the patient’s chart. The patient was asked to complete the second part of the satisfaction questionnaire before dismissal, and the physician was asked to complete the provider satisfaction questionnaire. The woman’s cytology and histology results were processed and read by the site’s routine pathology laboratory and were entered into the study database. The review site colposcopist was blinded to the primary colposcopist’s impressions, the woman’s cytology and histology, and any information on the questionnaires.

Statistical Analysis

Our study has a 90% power to determine a statistically significant k correlation, at the 0.05 2-tailed level, between the remote and review site colposcopists’ identification of the SCJ and their impressions of lesion severity and between each colposcopist’s impression and pathology. We used k correlation, independent and paired t tests, and descriptive statistics in this study. All statistics were calculated with Statistica (StatSoft, Tulsa, Oklahoma).²⁴

Results

A total of 79 women participated in our study. Sites 1 and 2 captured images of 39 and 40 women, respectively, representing 100% of the women scheduled for colposcopy at each site during the study. All women were white and ranged in age from 18 to 57 years, with a mean age of 29.5 years (standard deviation=11.5). Thirty-five percent of the women received their colposcopy at their usual site of gynecologic health care; the others were referred from outside offices where colposcopy was not available. The [Table] shows the descriptive data of the women participating in the project: 79% had a history of previous abnormal anogenital lesions; 65% presented with either atypical squamous cells of undetermined significance or low-grade squamous intraepithelial lesion (LSIL) cytology, and approximately 20% had a CIN 2/3 lesion histologically.

Technical Feasibility

The number of images captured for each patient varied from 3 to 20, and the details of the patient’s gynecologic history varied from the current visit data to an extensive past medical and gynecologic history. The average time for transmission of one patient’s images was 6 minutes.

Technical feasibility was measured by the percentage of agreement between the review and both remote sites for colposcopic impressions and histology correlates. A k value greater than 0.75 denotes excellent agreement, between 0.4 and 0.75 is good agreement, and below 0.4 is marginal agreement.25 For all examinations, the physician at the remote sites attempted to capture the SCJ for transmission. The SCJ was noted as “completely visualized” or “not visualized.” There was 86% agreement in complete visualization of the SCJ between the review and the remote sites (k=.533; 95% confidence interval [CI], 0.276-0.791; P=.019).

Both remote and review site physicians classified the colposcopic cervical impression as normal, LSIL, or high-grade-squamous intraepithelial lesion (HSIL). The interobserver agreement for colposcopic impressions was 86% (k=.684; 95% CI, 0.544-0.825; P <.001), and 66% for colposcopic impressions with exact histology (k=.416; 95% CI, 0.323-0.509; P <.001). If agreement between colposcopic impression and histology was relaxed to within 1 degree of severity, there was an 86% interobserver correlation (k=0.78; 95% CI, 0.697-0.861; P <.001). The distinction between normal metaplastic epithelium and low-grade changes accounted for all the discrepancies.

In our study, the implementation of a telecolposcopy system within the current office structure was easily accomplished, and most women accepted the telecolposcopic examination.* A total of 90.5% of the women surveyed travel 25 miles or less to their usual site of general health care; 88.1% traveled 25 miles or less for their colposcopy appointment. When the women were asked to choose whether they would prefer to have a colposcopy done locally and have the images electronically sent to another physician to be reviewed or drive farther than usual to the nearest experienced colposcopist, 95.2% of the women stated that they would have their colposcopy locally with electronic submission if the experienced colposcopist were more than 25 miles away. For the 12% of women who traveled more than 25 miles for their scheduled colposcopy, all stated they would have colposcopy locally if the experienced colposcopist was farther away than their current colposcopist.

Discussion

The telecolposcopy system captured the SCJ extremely well. Our study had an interobserver correlation of SCJ visualization similar to that seen for comparisons between different readers of the same cervigram image.^26,27 To compare between review and remote sites, the colposcopist at site 1 determined the impression while at the colposcope, and the colposcopist at the review center used the same patient’s digital image. The interobserver agreement for our colposcopic image versus digital image colposcopic impressions (86%) was higher than that when both colposcopic impressions were read from colpophotographs.²⁸ Our interobserver correlations between the colposcopic impression and histology agreed with others’ correlations for exact histologic agreement^29,30 (66%) references and was within 1 degree of disease severity (86%) reference number.³¹ The difference in interobserver agreements for the colposcopic impression and histology correlates reflects the difficulty in determining normal metaplasia from low-grade disease, a more difficult distinction than that between normal and high-grade or cancer disease.^8,28

Implementation of any organizational change within a medical office can often be difficult. It is more easily accepted when the organizational change is part of a clinical study.³² Both remote sites adopted the full telecolposcopic system (patient history data and image collection) after our study, because the teaching system, consultation reports, and record documentation were superior to the previous systems. Our study initially gave them the support and structure to use the telecolposcopic system. Implicit in adopting a new computerized system is the availability of someone to troubleshoot software and hardware problems. A remote telecolposcopic system requires that the main technical support comes from the review center, with onsite local technical help available, if necessary. In our study, technical help provided by the review center successfully kept the remote telecolposcopy systems functional with minimal effort.

More than two thirds of the women in our study were referred from their usual site of gynecologic health care to the colposcopist, and most traveled 25 miles or less to keep the appointment. Other studies have shown that women will drive more than 200 miles to participate in health care screening activities.³³ We have shown that when given the choice of a local provider with telecolposcopy access or driving more than 25 miles for an experienced examination, 95% of the women preferred to receive their care from a local colposcopist. This implies that for women in rural communities, local access to colposcopy with telecolposcopic links is preferred to traveling to a distant center of colposcopic excellence. Future efforts at increasing access to colposcopy for women should be directed toward local telecolposcopic connections with a regional center of excellence for image referral.

Conclusions

We have demonstrated that telecolposcopy is feasible, can be implemented in mountainous rural office sites, and is acceptable to the women undergoing the examination. To move from a demonstration project to a regional telecolposcopic network, telemedicine standards established for all telespecialties must be followed, including appropriate licensure and encryption standards.³⁴ Following these standards, the telecolposcopic network has the potential to offer increased access to and improved accuracy of colposcopic services. In addition, as an educational tool telecolposcopic images can be used for medical student, resident, and continuing medical education instruction, as well as direct patient education. Colposcopic training within an obstetrics and gynecology or family medicine residency training program occupies a small proportion of the curriculum. Telecolposcopy can be used to assist residents and other health care providers located at remote sites by providing access to experienced faculty and a large database of patients to review. The digital cervical images, if focused correctly, can provide more detail than the current hand-drawn documentation, which may lead to better cytologic, colposcopic, and histologic correlation.⁹

Colposcopists who as individuals perform a small number of colposcopies can gain knowledge from communicating between practices with the other single colposcopists through this network. They can all then anonymously compare their diagnoses with the actual biopsy result as a voluntary quality assurance and improvement program. This benchmarking can provide the groundwork for national colposcopic standards, as has been done in other countries^35,36 and in the United States for mammographic image and histology correlations.³⁷ The telecolposcopic system could focus on increasing access to colposcopy and on improving colposcopic pattern recognition skills so that high-grade disease would be targeted for treatment, and the overtreatment of low-grade disease would be avoided.

Acknowledgments

The New Hampshire Health Care Transition Fund Grant 98136, the Koop Institute, the Robert Wood Johnson Foundation, and the Division of Gynecologic Oncology, Dartmouth Medical School, have supported this work.